OPINION
BRUGGINK, Judge.Only the issue of damages remains in this Winstar-related1 case, which is on remand from the Federal Circuit. See LaSalle Talman Bank, F.S.B. v. United States, 317 F.3d 1363 (Fed.Cir.2003) (“LaSalle II"). Plaintiff, LaSalle Talman Bank, F.S.B. (“LaSalle”), offers two damages claims: lost profits and eost-of-replacement-eapital. A second trial was held. The matter has been fully briefed and orally argued. This court’s judgment of September 30, 1999, is vacated. For reasons set out below, we find that plaintiff’s lost *92profits claim was not established with reasonable certainty but award $8,288,700 for damages previously determined and not precluded by rejection of the current claim. As for the cost-of-replacement-capital claim, we accept plaintiffs model with two exceptions, which require us to remand the matter to the parties for calculation of the correct amount of recovery.
PROCEDURAL HISTORY
In an earlier decision, we found the government hable for breach of contract following enactment and implementation of the Financial Institutions Reform, Recovery and Enforcement Act of 1989 (“FIRREA”), Pub.L. No. 101-73, 103 Stat. 183 (Aug. 9, 1989) (codified in scattered sections of 12 U.S.C.). Cal. Fed. Bank, F.S.B. v. United States, 39 Fed.Cl. 753, 765-66, 779 (1997) (“Cal.Fed.I”) (including consolidated case LaSalle Talman Bank, F.S.B. v. United States, 92-652C). A trial was then held on plaintiffs damages claims. LaSalle Talman Bank, F.S.B. v. United States, 45 Fed.Cl. 64 (1999) (“LaSalle I"). In LaSalle I we rejected most of the government’s defenses. We nevertheless denied ah of plaintiffs damages claims based on restitution, lost profits, and cost-of-replacement-eapital theories. We awarded only $5,008,700 in incidental damages for expenses incurred in connection with the FIR-REA-foreed sale of LaSalle’s predecessor bank, Talman Home Federal Savings and Loan Association of Illinois (“Talman”), to ABN AMRO N.A., Inc. (“ABN AMRO”). Id. at 120.2
On appeal the Federal Circuit affirmed the judgment of liability for breach of contract. LaSalle II, 317 F.3d at 1370. Our rejection of the government’s defenses was upheld. The denial of plaintiffs restitution claim was also upheld, although on somewhat different grounds. Id. at 1376-77. Our disposition of the lost profits and cost-of-replacement-capital claims, however, was not affirmed.
The Federal Circuit confirmed that plaintiffs lost profits claim could be an appropriate means for determining FIRREA-induced damages. Id. at 1370-71. It also agreed that lost profits must be adjusted to account for earnings directly attributable to ABNAMRO-financed investments in the thrift that would not have occurred but for the breach. Id. The offset found in LaSalle I, however, included earnings on ABN AMRO investments that were not directly attributable to the breach. LaSalle I, 45 Fed.Cl. at 91. Our original offset, in fact, would have fully mitigated the lost profits identified at the time. Id. The Federal Circuit, however, concluded that the offset must be limited to earnings on ABN AMRO’s initial $300 million capital infusion. LaSalle II, 317 F.3d at 1374. Consequently, our holding was vacated and the issue remanded so that a properly limited calculation of the earnings offset could be made. Id. at 1374.
The Federal Circuit also rejected our ultimate holding on the cost-of-replacement-eapital damages claim. Under this theory, it was argued that the $300 million infusion, which partially mitigated the disallowance of plaintiffs supervisory goodwill, came at a price: dividends paid by plaintiff to ABN AMRO. We had held that such costs were not recoverable because plaintiffs obligation to pay was not legally enforceable. LaSalle I, 45 Fed.Cl. at 112. The Federal Circuit reversed, holding that the cost of capital is not dependent on whether an infusion triggers a legal obligation to pay dividends. LaSalle II, 317 F.3d at 1375. The issue was remanded for determination of the dividends attributable to plaintiffs mitigation, less any value gained by holding cash rather than supervisory goodwill. Id.
On remand, plaintiff once again pursues both lost profits and eost-of-replaeement-capital damages theories. Plaintiffs specific damages models, however, differ substantially from those offered at the initial trial and argued on appeal. In the following section, we summarize only those facts pertinent to the theories at issue on remand.3
*93FACT BACKGROUND
Faced with its own troubled future in the midst of the savings and loan crisis of the early 1980s, Chicago-based Talman agreed to supervisory mergers with four failing thrifts in its own market in 1982. LaSalle I, 45 Fed.Cl. at 71-72. As a result, Talman assumed $912.6 million in net liabilities from the merged thrifts. Id. at 72. Under the regulatory scheme that encouraged these mergers, Talman was permitted to count the new net liabilities as supervisory goodwill, an intangible asset that existed only on paper. Id. Talman’s newly-acquired goodwill was to amortize over forty years. Id. at 71.
Talman’s management followed a conservative business strategy that emphasized single-family mortgage lending funded by consumer deposits. This business strategy enabled Taiman to return to profitability by 1986. Id. During that year, Taiman converted from a mutual association to a shareholder-owned association. Id. at 72. As part of Talman’s conversion agreement with the Federal Savings & Loan Insurance Corporation (“FSLIC”), Talman’s forty-year supervisory goodwill amortization period was reduced to thirty years. Id. Taiman operated profitably from 1986 until the enactment of FIRREA in 1989. Taiman earned net profits of $28 million in 1986, $21 million in 1987, and $26 million in 1988. Talman also continued to experience growth through 1989.
In August 1989, Congress enacted FIR-REA. Prior to FIRREA, Taiman could count the supervisory goodwill on its books toward the minimum level of capitalization mandated by financial regulations. FIRREA called for the phase-out of the use of supervisory goodwill by December 31, 1994. The Office of Thrift Supervision (“OTS”) issued implementing regulations in November 1989 that phased out supervisory goodwill by the end of 1993. At the time of FIRREA’s enactment, Taiman held $514 million in supervisory goodwill. The elimination of this supervisory goodwill seriously disrupted Talman’s business operations. Talman’s inability to count supervisory goodwill on its books threatened to render the thrift insolvent and force it out of compliance with regulatory capital requirements.
At this point, Taiman was in danger of being placed into receivership by the Resolution Trust Corporation (“RTC”). RTC only permitted the thrift’s survival through regulatory forbearance. Taiman was required to submit a capital plan for OTS approval. On March 16, 1990, OTS approved a plan containing two critical conditions: Taiman was prohibited from making dividend payments to its shareholders, and the deadline for achieving capital compliance was brought forward to December 31,1993 (one year earlier than the deadline established by FIRREA). Among the immediate consequences of FIR-REA’s enactment and the plan’s implementation was the need to shrink Talman’s asset portfolio in an effort to make it compliant with capital regulatory requirements. Taiman shrank its portfolio $1 billion dollars between September 1991 and March 1992.
The enactment of FIRREA prompted Talman’s retention of Salomon Brothers, Inc. to serve as its investment banker and to investigate possible recapitalization methods. Two possibilities investigated were a merger with another financial institution that could provide Taiman with a capital infusion and a stock issue that could raise capital. Salomon’s efforts were successful. By early 1991 Taiman was engaged in discussions with both ABN AMRO, a large multi-national bank based in the Netherlands, and Bank of America Corp. Ultimately, Talman’s board voted on July 15, 1991, to accept ABN AMRO’s offer to purchase the thrift’s outstanding stock at $10 per share.
When the acquisition was completed on February 28,1992, ABN AMRO owned 100% of Talman’s stock. The stock cost ABN AMRO $97 million. As part of the merger agreement, Taiman agreed to dispose of particular assets. Specifically, Taiman sold certain mortgage-backed securities for $42.1 million. Another agreement term, mandated by OTS, required ABN AMRO to infuse Taiman with $300 million in capital.
Upon acquisition, Taiman ceased to exist and LaSalle came into being. The new thrift’s assets and liabilities were marked to market because the acquisition employed purchase accounting. This had two immediate effects: first, all remaining supervisory *94goodwill was eliminated because the market value of LaSalle’s assets exceeded their book value; and second, when combined with the $300 million infusion, LaSalle’s core leverage ratio was boosted to 6.65%, bringing LaSalle into compliance with capital requirements. Taiman would have held $463.9 million in unamortized supervisory goodwill at the time of acquisition.4 After FIRREA and the acquisition, ABN AMRO’s $300 million infusion left LaSalle with $163.9 million of unmitigated, disallowed supervisory goodwill.
As a consequence of the acquisition, LaSalle held $411 million in paid-in capital. This consisted of the $300 million infusion and purchase-accounting adjustments of $111 million. At this point, LaSalle again began to grow.
DISCUSSION
I. Lost Profits Damages
A. The Current Lost Profits Model
On remand, plaintiff offers a lost profits model that differs greatly from the model initially advanced during this litigation. At the first trial, plaintiffs expert witness, Prof. Christopher James, Ph.D.,5 constructed a lost profits model that compared the profits of two hypothetical banks: the “But-For Bank” and “Old Talman.” LaSalle I, 45 Fed.Cl. at 89. The But-For Bank represented profits that would have been earned by Talman in a hypothetical world in which there was no breach. Plaintiff argued that the But-For Bank’s earnings should include: foregone earnings on a merger with Cragin Federal Bank for Savings (“Cragin”) that was delayed due to the breach; mortgage servicing earnings foregone between 1990 and 1992; profits lost due to the $1 billion shrink in assets; and wounded-bank damages. Old Talman, on the other hand, was a reflection of Talman in a hypothetical world in which it failed to secure a purchase by ABN AMRO and the attendant $300 million capital infusion. Plaintiff argued that Old Talman’s earnings should be adjusted down to account for actual earnings resulting from the 1994 Home Savings Bank (“Home Savings”) acquisition and the 1995 Cragin merger-conversion. Asserted lost profits consisted of the difference between the earnings of the BuLFor Bank and those of Old Taiman — $858.8 million.
After trial, but prior to ruling, the court proposed an alternative lost profits formulation in an effort to arrive at a settlement figure acceptable to both parties. This proposal projected earnings on assets LaSalle might have held but for the breach. Following the court’s suggestion and drawing raw figures from the existing record, plaintiff submitted a document detailing a projection of earnings lost on foregone assets. This projection presumed that LaSalle, absent the breach, would have leveraged its supervisory goodwill at a 7% rate. Plaintiff calculated that tens of billions of dollars of assets were foregone between the breach and 2012, the year when supervisory goodwill would have been fully amortized. A 0.8% rate of return was applied to this pool of foregone assets. According to the projection, plaintiff lost $401.3 million in earnings on assets foregone due to the breach.
The parties were unable to settle. We therefore confined our ruling to the merits of the lost profits model developed during trial. We held that Old Taiman was an inappropriate point of comparison for the Butr-For Bank. We ruled that the actual bank’s earnings, which were much higher than those of Old Taiman, should have been used. We also ruled that a number of upward adjustments to the But-For Bank’s earnings were either partially or fully unsupported. Ultimately, we concluded that plaintiff had incurred $13.5 million in lost profits and incidental damages as a result of the government’s breach. The lost profits portion of our finding, however, was offset by plaintiff’s earnings on several capital infusions made by ABN AMRO throughout the 1990s. Therefore, we only *95awarded plaintiff incidental damages of approximately $5 million.
On appeal, plaintiff chose to tack a different course from the one pursued at trial. Plaintiff did not challenge any of the adverse rulings on adjustments it sought under its first lost profits model. Plaintiff only challenged the size of the offset applied by this court. Had this been plaintiffs sole argument, it would have proved fruitless regardless of its success. If the offset calculation were limited to ABN AMRO’s $300 million infusion, earnings on the cash likely would have been greater than the lost profits previously found. Therefore, in conjunction with its offset argument, plaintiff advanced a second lost profits model on appeal that basically adhered to the outline developed in connection with the post-trial settlement discussion, projecting $401.3 million in lost profits.
On remand, plaintiff has again altered its lost profits model. The current model melds the second model’s earnings-on-foregone-assets methodology with the damages identified in our first opinion and the Federal Circuit’s holding concerning the offset. Plaintiff limited the projection to five years, adjusted the balance of supervisory goodwill downward in response to ABN AMRO’s initial $300 million infusion, and tailored the result to account for the damages found in LaSalle I.
Using this approach, plaintiff now projects $31.2 million in earnings lost on assets foregone between March 1, 1992, and December 31,1996 (“foregone-asset period”). The foregone-asset period begins the day after the ABN AMRO acquisition and ends when LaSalle first returned capital to ABN AMRO. The projection computes lost profits on an annual basis. Although plaintiff would have held $463.9 million in supervisory goodwill on the first day of the foregone-asset period, plaintiff tailored the projection to account for the mitigating effect of the $300 million capital infusion. Therefore, the projection begins with a $163.9 million balance of unmitigated supervisory goodwill.6 Each year’s balance is adjusted downward to account for the thrift’s uncontested tangible capital deficiency. Furthermore, the balance declines throughout the foregone-asset period in response to amortization. The balance of adjusted supervisory goodwill therefore declines from $106.9 million to $12.9 million over the course of the foregone-assets period.
The earnings-on-foregone-assets projection employs a 7% leverage ratio, which is an approximation of the thrift’s actual core capital ratio throughout the foregone-asset period.7 Each year’s average balance of supervisory goodwill is multiplied by this leverage ratio to determine the amount of foregone assets. The product is then reduced by that year’s average balance of adjusted supervisory goodwill to account for regulatory capital minimums. What remains is a projection of assets foregone in any given year. Foregone assets decline from $1.2 billion in December 1992 to $293.1 million in December 1996.
Each year’s projected foregone assets are then multiplied by LaSalle’s actual pre-tax, pre-amortization return on average assets (“ROAA”).8 This computation yields the annual earnings allegedly lost due to the breach. In total, plaintiff projects $31.2 million in earnings were lost over the course of the foregone-asset period.
Plaintiff added a component to its lost profits model to account for our findings in LaSalle I that plaintiff had suffered $13.5 million in incidental and lost profits damages. We found that plaintiff had incurred $5 million in wounded-bank damages, $1.9 million in foregone-mortgage-servicing damages, and $6.6 million in lost profits due to the shrink of its asset portfolio. LaSalle I, 45 Fed.Cl. at 99. We only awarded $5 million for incidental wounded-bank damages, finding that *96the remaining lost profits damages were fully offset by profits attributable to the ABN AMRO acquisition.
Our $5 million incidental damages award was upheld by the Federal Circuit. LaSalle II, 317 F.3d at 1377. Plaintiff therefore adds this amount to its lost profits model. Our findings concerning lost profits were not disturbed by the Federal Circuit, but our decision concerning the offset was vacated and remanded for recalculation. Id. at 1374. Plaintiff contends that its new lost profits model satisfies the Federal Circuit remand. The only new capital that must be accounted for after the remand is the $300 million infusion. The need to account for other capital infusions was rejected by the court of appeals. Rather than directly account for the earnings on the $300 million, however, plaintiff proposes to do so indirectly by excluding from its claim anything other than earnings lost on the unmitigated supervisory goodwill, i.e., supervisory goodwill in excess of $300 million. According to plaintiff, earnings on the initial ABN AMRO infusion are irrelevant in the new model. It therefore adds our $1.9 million foregone-mortgage-servicing finding and a portion of our $6.6 million shrink finding to its lost profits model.
Plaintiffs claim for lost profits remains complicated, however, by an overlap in time between the foregone-asset period and the previous claim for earnings lost due to the involuntary $1 billion shrink in assets. The court previously accepted the argument that the bank lost money because of the shrink but found that the shrink began in September 1991 and ended in September 1993. LaSalle I, 45 Fed.Cl. at 95, 96. We found that the plaintiff lost $6.6 million because of that shrink but declined to award it because of our now-reversed finding with respect to offsetting earnings. Plaintiffs current claim seeks to have the earlier finding of shrink damages reinstated, at least to the extent that it does not overlap with the current claim for earnings on foregone assets. The non-overlapping amount, according to plaintiff, is $1.38 million, which represents profits lost due to the shrink prior to the foregone-asset period.
B. The Law of the Case Doctrine
Defendant contends that the plaintiffs lost profits model is barred by the law of the case doctrine because the model’s main component, the earnings-on-foregone-assets projection, is being pursued for the first time on remand. This judicially created doctrine limits the revisitation of issues decided earlier in the same litigation. Exxon Corp. v. United States, 931 F.2d 874, 877 (Fed.Cir.1991). The doctrine encourages a judge’s adherence to previous trial-level pronouncements and mandates adherence to any appellate-level determinations. Id. The latter dimension of a trial judge’s obligation is frequently referred to as the “mandate rule”9 — a judge presiding over a case on remand is bound to adhere to the appellate court’s mandate on all issues that were either explicitly or implicitly addressed by the reviewing court. See 18 James Wm. Moore et al., Moore’s Federal Practice § 134.23[1] (3d ed.2000); see also Ins. Group Comm. v. Denver & Rio Grande W. R.R., 329 U.S. 607, 612, 67 S.Ct. 583, 91 L.Ed. 547 (1947).
In LaSalle II, the Federal Circuit’s holding with regard to plaintiffs first lost profits model was narrow: ‘We vacate the decision that no damages accrued on the lost profits model, and remand for calculation of the profits attributable to the $300 million capital investment, to be credited against Talman’s projected lost profits.” 317 F.3d at 1374. This holding makes two points: first, our ultimate determination awarding zero lost profits damages is vacated; and second, the offset must be recalculated and credited against Talman’s lost profits. The re-evaluation of the offset is the crux of the Federal Circuit’s opinion. In its absence, no other reversible error warrants the remand. The amount of lost profits, for example, was not considered by the Federal Circuit. The *97mandate rule’s effect is thus limited to the offset issue because no other issue was challenged on appeal. Therefore, our re-examination of plaintiffs loss, independent of the offset, is not barred by the terms of the mandate.
The mandate rule is but a single facet of the broader law of the case doctrine, however. Whereas the mandate rule places an affirmative duty on a trial judge to adhere to an appellate ruling, the overall doctrine’s application is discretionary with regard to prior trial-level determinations of an issue. According to the Federal Circuit, the law of the case doctrine “has long been held not to require the trial court to adhere to its own previous rulings if they have not been adopted, explicitly or implicitly, by the appellate court’s judgment.” Exxon, 931 F.2d at 877. Although the Exxon court acknowledged that, for the sake of “[o]rderly and efficient ease administration,” decided questions should not be revisited in all circumstances, it explicitly preserved a trial court’s right to reconsider its own decisions prior to the entry of judgment. Id.
The law of the case doctrine thus does not preclude our examination of plaintiffs newest lost profits model because the model presents a method for measuring lost profits that was not directly pursued at the first trial. As plaintiff characterizes its claim, the eamings-on-foregone-assets theory is simply a method for ascertaining the earnings lost on unmitigated supervisory goodwill. The theory is based on the assumption that the thrift would have fully and profitably leveraged every dollar of its missing goodwill into profitable assets. As the court understands the underlying assumption, plaintiff is not limited by the absolute dollar amount of any historical shrink in assets, although that shrink is, of necessity, embedded within the claim.
Defendant argues, however, that two specific findings in LaSalle I necessarily address and preclude plaintiffs newest lost profits model. First, we found that the thrift’s portfolio had regained its original size two years after it began its $1 billion shrink. We therefore limited plaintiffs damages to profits on assets lost in the shrink that had yet to be replaced before the thrift’s portfolio had regained its pre-shrink size. Defendant characterizes our finding as preclusive of plaintiffs current earnings-on-foregone-assets projection.
Plaintiff offers us no reason to re-examine our finding that the shrink ended in two years. It contends, in substance, that the finding holds no direct consequences for its current theory, presumably because the recovery of the portfolio’s size does not differentiate between the sources of asset growth. The response is thus, “that was a different theory; we are entitled to write on a clean slate after the remand.” As we understand the new claim, plaintiff projects assets that would have been held but for the breach, irrespective of whether any particular asset can be characterized either as having been eliminated due to the shrink or as simply never having been acquired. The new theory thus focuses on foregone assets on the assumption that, irrespective of how large the portfolio grew, it could have grown larger with more capital. Plaintiff therefore argues that, so long as shrink damages are not double counted, the new claim can proceed.
We will permit plaintiff to proceed. Therefore, plaintiffs earnings projection is not barred by the limited duration of the shrink. We acknowledge, however, that the continuing evolution of theories at some point runs the risk of becoming problematic. To the extent the claim reflects the historic shrink in the portfolio, it is partially resolved in the prior ruling. To the extent it projects assets never held, it becomes inherently more difficult of proof, as we discuss below.
Second, defendant points to our observation in LaSalle I that plaintiff failed to pursue a foregone-asset theory at trial. 45 Fed. Cl. at 102. Defendant ascribes inappropriate significance to this comment. The absence of an earnings-on-foregone-assets argument was merely noted in support of our conclusion that the offset’s size greatly outweighed damages — it was not a finding on the merits of the foregone-asset theory.
In sum, the Federal Circuit’s mandate only constrains our analysis of the offset. It does not impinge on our re-examination of other damages issues. As to the scope of our *98previous holding, it remains within our discretion to re-examine the amount of profits lost by plaintiff. We are free, therefore, to weigh the merits of plaintiffs newest lost profits model.
C. Expectancy Damages
It is a fundamental precept of contract damages that a plaintiff is entitled to those contract benefits it reasonably expected, but for the breach. LaSalle II, 317 F.3d at 1371. A plaintiff seeking expectancy damages must establish by a preponderance of the evidence that 1) “the loss of profits caused by the breach was within the contemplation of the parties because the loss was foreseeable or because the defaulting party had knowledge of special circumstances at the time of contracting”; 2) “the loss was the proximate result of the breach”; and 3) “a sufficient basis exists for estimating the amount of lost profits with reasonable certainty.” Energy Capital Corp. v. United States, 302 F.3d 1314, 1325 (Fed.Cir.2002) (citations omitted); see also Bluebonnet Sav. Bank, F.S.B. v. United States, 266 F.3d 1348, 1355 (Fed.Cir.2001).
1. Foreseeability
Damages are foreseeable if they “ ‘follow[ ] from the breach of contract in the ordinary course of events.’ ”10 Home Sav. of Am., F.S.B. v. United States, 57 Fed. Cl. 694, 726 (2003) (quoting Restatement (Second) Contracts § 351). Both this court and the Federal Circuit have concluded in the context of Winstar litigation that government regulators expected thrifts to profit from the supervisory goodwill on their books. See, e.g., Cal. Fed. Bank, F.S.B. v. United States, 245 F.3d 1342, 1349-50 (Fed.Cir.2001) (“Cal.Fed.II”); Commercial Fed. Bank, F.S.B. v. United States, 59 Fed.Cl. 338, 354-55 (2004); Home Sav., 57 Fed.Cl. at 727. As the Supreme Court observed in Winstar, pre-FIRREA treatment of supervisory goodwill was “attractive because it inflated [an] institution’s reserves, thereby allowing the thrift to leverage more loans (and, it hoped, make more profits.)” 518 U.S. at 851, 116 S.Ct. 2432.
We reached the same conclusion in LaSalle I. It was clear from testimony and government documents that regulators expected Taiman to leverage its supervisory goodwill for a profit.11 We concluded on that basis that the disallowance of supervisory goodwill would lead to a foreseeable loss of profits. Id. The second trial has not altered our view on this issue. The possibility that the disallowance of supervisory goodwill would result in lost profits was foreseeable.
2. Causation
Plaintiff bears the burden of proving that its damages were the proximate result of the government’s breach. See Energy Capital, 302 F.3d at 1325. The Federal Circuit recently re-addressed the question of causation in connection with a Winstar lost profits claim in California Federal Bank v. United States, the latest opinion from the California Federal litigation. 395 F.3d 1263 (Fed.Cir.2005) (“Cal.Fed.IV”). California Federal Bank’s (“CalFed’s”) claim against the United States was similar to LaSalle’s in that it sought profits lost on an asset shrink undertaken in response to FIRREA. In an earlier opinion addressing the government’s successful motion for summary judgment on the lost profits claim, the Federal Circuit remanded the question of lost profits. Cal. Fed. II, 245 F.3d at 1342. It held that lost profits are appropriate when they are “ ‘the proximate result of the breach’ ” and their but-for existence is definitely established. Id. at 1349 (quoting Neely v. United States, *99152 Ct.Cl. 137, 285 F.2d 438, 443 (1961)). The court went on to make the following distinction concerning proximate causation: profits that would have accrued as “‘the direct and immediate result[ ]’ ” of the contract’s fulfillment are recoverable; profits that would have been realized by “ ‘independent and collateral undertakings’” are not. Id. (quoting Wells Fargo Bank, N.A. v. United States, 88 F.3d 1012, 1022-23 (Fed.Cir. 1996)).
The Federal Circuit concluded that the purpose of the parties’ agreement was to provide CalFed with favorable regulatory treatment in exchange for its absorption of a number of failing thrifts. Id. The court thus held that CalFed’s use of supervisory goodwill was a central focus of the agreement and the subject of the breach, and that “profits on the use of ... supervisory goodwill ... [were] recoverable as damages.” Id. The Federal Circuit made it clear that disallowance of supervisory goodwill caused the thrift to lose profits.
On remand, the trial court rejected the lost profits claim once again. Cal. Fed. Bank, F.S.B. v. United States, 54 Fed.Cl. 704 (2002) (“Cal.Fed.III”). Plaintiff appealed, challenging, among other aspects of the ruling, the court’s asserted error in rejecting a “substantial factor” test that plaintiff discerned in the prior appellate decision. In the most recent decision, however, the Federal Circuit makes clear that the substantial factor test is not appropriate: “the causal connection between the breach and the loss of profits must be ‘definitely established.’ ... [T]he Court of Federal Claims correctly rejected the ‘substantial factor’ test advocated by CalFed.” Cal. Fed. IV, 395 F.3d at 1268 (citations omitted).
As in the California Federal litigation, here the government granted Taiman favorable regulatory treatment in exchange for its participation in the Phoenix program. LaSalle I, 45 Fed. Cl. at 71-72. Taiman used its supervisory goodwill to comply with relevant regulations. Id. Talman’s use of supervisory goodwill, like CalFed’s, was a central focus of its agreement with the government and the subject of the breach. As we previously concluded, FIRREA forced Taiman, like CalFed, to shrink its asset portfolio. Id. at 95-96. Therefore, profits that would have been earned on assets lost due to the shrink are recoverable.
LaSalle’s lost profits claim seeks earnings in addition to the $6.6 million previously found for the shrink — earnings on assets it argues it was forced to forego because FIR-REA disallowed the use of supervisory goodwill as regulatory capital. Although CalFed pursued a similar claim on remand, it failed to appeal the claim’s denial. The Federal Circuit’s initial decision in California Federal II, however, retains its vitality with respect to the causal connection between loss of supervisory goodwill, foregone assets, and lost earnings.
The Federal Circuit accepted the contention that the breach caused CalFed to lose profits because the thrift was prevented from using goodwill as anticipated in its agreement with the government. The cause of Talman’s alleged foregone assets is no different. Absent the breach, supervisory goodwill would have enabled Taiman to avoid shrinking its asset portfolio and, potentially, enlarge its portfolio further with the unmitigated supervisory goodwill. There is no doubt that the disallowance of goodwill caused Taiman to reduce its operations beyond merely the period of the shrink.12 Whether LaSalle adequately proved with reasonable certainty the profits it lost on its foregone assets will be explored in the following section.
Although California Federal II and IV provide the Federal Circuit’s clearest pronouncements on the lost profits causation issue in the Winstar context, we would be remiss if we failed to discuss other eases arising in the same context. In Glendale Federal Bank, F.S.B. v. United States, for example, the Federal Circuit addressed the lost profits causation issue in dicta. 239 F.3d *1001374, 1380 (Fed.Cir.2001). Glendale Federal did not appeal the trial court’s rejection of its lost profits claim. The appellate court nevertheless commented that problems of proof in these cases make any recovery under a lost profits theory exceedingly difficult. Id. In a second Glendale appeal, the Federal Circuit took its commentary even further: “Expectancy damages theory ... [is] generally not susceptible to reasonable proof____Experience suggests that it is largely a waste of time and effort to attempt to prove such damages.” Glendale Fed. Bank, F.S.B. v. United States, 378 F.3d 1308, 1313 (Fed.Cir. 2004).
This commentary lacks the force of law, yet it brings into clear relief a legal hurdle that has proven difficult in the past for many Winstar plaintiffs in their pursuit of lost profits damages. See, e.g., Bank United v. United States, 80 Fed.Appx. 663 (Fed.Cir. 2003); Columbia First Bank, F.S.B. v. United States, 60 Fed.Cl. 97, 107-08 (2004); S. Cal. Fed. Sav. & Loan. Ass’n v. United States, 57 Fed.Cl. 598, 626 (2003); Cal. Fed. III, 54 Fed.Cl. at 715. The commentary in the Glendale opinions has been relied upon in subsequent decisions by the Federal Circuit and this court. See, e.g., Cal. Fed. IV, 395 F.3d 1263, 1271-72; Standard Fed. Bank v. United States, 62 Fed.Cl. 265, 273 (2004); Columbia First, 60 Fed.Cl. at 102. Other decisions, while not citing Glendale, rely on a similar rationale. See S. Nat’l Corp. v. United States, 57 Fed.Cl. 294, 305-06 (2003); Fifth Third Bank of W. Ohio v. United States, 55 Fed.Cl. 223, 228 (2003); Bank United of Tex. F.S.B. v. United States, 50 Fed.Cl. 645, 654-55 (2001).
In the final analysis, however, “both the existence of lost profits and their quantum are factual matters” that are to be decided by an examination distinct from the one used to determine causation. Cal. Fed. II, 245 F.3d at 1350. Here, as in California Federal, lost profits are a proximate result of the government’s breach. Whether plaintiff can prove a basis for estimating lost profits with reasonable certainty remains to be seen.
3. Reasonable Certainty
Under the final prong of the standard outlined in Energy Capital, LaSalle must provide a sufficient basis for proving its damages with reasonable certainty. 302 F.3d at 1325. The Federal Circuit has also cautioned that, when a “reasonable probability of damage can be clearly established, uncertainty as to the amount will not preclude recovery.” Cal. Fed. II, 245 F.3d at 1350 (quoting Locke v. United States, 151 Ct.Cl. 262, 283 F.2d 521 (1960)); see also LaSalle II, 317 F.3d at 1374. Plaintiffs damages thus need only be founded upon a “reasoned conclusion.” See Palmer v. Conn. R. & Lighting Co., 311 U.S. 544, 561, 61 S.Ct. 379, 85 L.Ed. 336 (1941).
Plaintiff argues that its earnings-on-foregone-assets projection is reasonably certain because it extrapolates LaSalle’s actual performance during the foregone-asset period. According to plaintiff, projecting LaSalle’s performance accurately approximates Talman’s performance in the but-for world because both entities shared Talman’s business strategy. In response, defendant challenges the model’s reliability on a number of grounds: 1) it does not account for LaSalle’s excess capital and high leverage ratio during the foregone-asset period; 2) it fails to identify specific foregone assets; 3) it fails to account for the law of diminishing marginal returns; and 4) its predicted rate of return is unsubstantiated.
Plaintiffs lost profits model projects $31.2 million in earnings on foregone assets. This represents earnings allegedly lost during the foregone-asset period on assets either disposed during the shrink of plaintiffs portfolio or foregone as a result of the breach. As explained in Part I.A, the earnings-on-foregone-assets projection is comprised of two calculations. Plaintiff first applies a 7% leverage ratio to disallowed supervisory goodwill in order to determine the amount of foregone assets. Plaintiff then applies the ROAA to these foregone assets to yield lost earnings. Both the leverage ratio and the ROAA are extrapolated from LaSalle’s actual performance during the foregone-asset period. Defendant challenges the accuracy of both calculations.
In previous cases, this court has held that Winstar plaintiffs seeking lost profits on *101foregone or shrunk assets must produce evidence of the types and relative proportions of assets that would have been held in the but-for world. See, e.g., Commercial Fed., 59 Fed.Cl. at 349 n. 29 (plaintiff offered evidence of specific opportunities in support of lost profits award); Columbia First, 60 Fed. Cl. at 115-17 (lost profits denied, in part, due to failure to specify the proportion of various assets in plaintiffs but-for portfolio); S. Nat’l, 57 Fed.Cl. at 306 (recovery barred due to failure to specify foregone assets).
In Columbia First, this court deemed overly speculative an earnings-on-foregone-assets projection similar to the one at issue here. 60 Fed.Cl. at 118. Columbia First’s projection flowed from a hypothetical business strategy that differed from the one actually employed by Columbia First. Id. at 114-115. The use of a model that relied on untested presumptions and strategies was too speculative to establish lost profits with reasonable probability. Id.; see also S. Cal. Fed., 57 Fed.Cl. at 626 (uncertain model’s rate of return was an average of other thrifts’ rates, not plaintiff’s). By extrapolating from LaSalle’s actual experience, plaintiff tailored its earnings-on-foregone-assets projection to address the specific concern raised in Columbia First.
Taiman operated profitably in the years prior to FIRREA, earning roughly $25 million a year in net profits between 1986 and 1988. Its long-term business strategy called for the mediation of consumer deposits into residential mortgages. At times when the residential loan market lacked strength, however, the strategy called for investment in mortgage-backed securities, purchases that were funded with federal loans. This short-term approach earned plaintiff a relatively modest profit until long-term opportunities again became available.
As Mr. Theodore Roberts, the thrift’s former chairman, explained, the thrift adhered to Talman’s business strategy after both the breach and the ABN AMRO acquisition. Mr. Harry Tempest13 concurred, testifying that “[i]t was our intention to follow the ... Taiman model for both retail banking and for mortgage banking and use that as our model for our retail activities throughout the Chicago enterprise.” Tr. at 492-93. For example, upon its merger with Cragin, LaSalle abandoned Cragin’s most profitable lending strategies, preferring instead to emphasize the types of stable assets traditionally pursued by Taiman. Plaintiffs Strategic Plan from 1994 to 1996 also demonstrates the thrift’s commitment to Talman’s core business strategy during the foregone-asset period. Such evidence is probative of lost profits in a Winstar case. See Bank United, 80 Fed. Appx. at 667 (holding that formal business plans are indicators of a thrift’s actual business strategy); see also Cal. Fed. II, 245 F.3d at 1350; Commercial Fed., 59 Fed.Cl. at 338-39.
We therefore find that LaSalle’s leverage ratio and rate of return on assets during the foregone-asset period offer a sensible way to initiate a projection of Talman’s potential loss of profits during the period in question. They are grounded in reality and based on Talman’s business model. Whether plaintiff fully accounts for other relevant factors is discussed below.
Calculation of the alleged amount of foregone assets is the first component of the projection. The mechanics of this calculation are straightforward and uncontested. The application of a 7% leverage ratio14 to a given year’s balance of average adjusted supervisory goodwill15 yields the value of assets foregone that year. Although the leverage ratio remains constant from year to year, the supervisory goodwill balance declines in *102response to amortization. Therefore, this calculation projects foregone assets from a high of $1.3 billion in 1992 to a low of $293 million in 1996.
Although its mechanics are sound, the calculation’s validity hinges on the availability of unutilized investment opportunities during the foregone-asset period. Plaintiff assumes that investment opportunities were in such abundance that Taiman would have had no difficulty leveraging its supervisory goodwill into profitable assets. Defendant counters with evidence that, far from being limitless, investment opportunities were scarce and could not have supported the projected amount of assets, at least at the assumed rate of return.
Defendant points to internal documents indicating that the thrift’s management was concerned that there were not sufficient long- and short-term investment opportunities during the foregone-asset period. Investment opportunity scarcity was referenced by members of the thrift’s management team at least six times between 1990 and 1993. These concerns are represented in the minutes of LaSalle’s Asset/Liability Management Committee (“ALCO”) and in correspondence between LaSalle’s management aixd OTS. When Mr. Roberts and Mr. Scott Heitmann16 were questioned about these documented statements at trial, each conceded that Talman’s investment opportunities were scarce at the time the statements were made. Tr. at 232, 240 (Roberts) (“There were periods where you had a dearth of [quality mortgage product]”); Tr. at 333 (Heitmann) (“[We had] a situation in the economy where loan growth or the px'ospects for loan growth [were not] very good.”).
Plaintiff points out, however, that, of the six references to scarce opportunities recorded in ALCO meeting minutes, only two occurred during the foregone-asset peiiod and both were early in the period. In response to this scarcity, the members shifted the thrift’s priorities to mortgage-backed securities. Tr. at 332 (Heitmann) (“[W]e would have then gone on to a discussion about what can we do to releverage our capital here____ [W]e were probably looking at buying securities in some form.”). This priority shift, as discussed above, was consistent with the Taiman business strategy.
Defendant also points to the many instances during the foregone-asset period in which ALCO members referred to LaSalle’s “excess capital,” suggesting that LaSalle held capital in excess of regulatory minimums because investment opportunities were scarce. As Mr. Heitmann testified, however, ALCO members identified unused capital so that it could be quickly paired with an unutilized investment opportunity. As loans were paid and earnings on securities collected, plaintiff was presented with unused capital that had to be redeployed in the market as quickly as possible. ALCO members referi’ed to unused capital as excess capital — the designation was not necessarily an indication of limited investment opportunities. The ongoing presence of some unused capital was inevitable because the composition of plaintiff’s portfolio was in constant flux.
We also note that, although ALCO members consistently discussed the thrift’s excess capital throughout the lost profits period, the thrift profitably grew its portfolio. Plaintiff was able to find investment opportunities despite holding capital in excess of regulatory minimums. We do not, however, push this evidence beyond its natural conclusion— just because plaintiff located profitable opportunities and grew throughout the foregone-asset period does not prove that plaintiff could have simultaneously replicated these results with lost supervisory goodwill in the but-for world.
Defendant also argues that ABN AMRO’s failure to infuse plaintiff with additional capital during the foregone-asset period proves that investment opportunities were not available. Although it concedes that ABN AMRO made a number of infusions during the lost profits period,17 defendant argues that ABN *103AMRO would have infused more capital if the thrift lacked enough to pursue all available investment opportunities. It is uncontested that ABN AMRO regularly down-streamed capital to its subsidiaries so that investment opportunities could be pursued. As Mr. Tempest testified, however, ABN AMRO, like all businesses, possessed a limited supply of capital. Only subsidiaries with the most profitable investment opportunities warranted a capital infusion.18 ABN AMRO’s failure to infuse plaintiff with additional capital suggests that, if investment opportunities had been available to plaintiff during the foregone-asset period, they were less profitable than the opportunities available to other ABN AMRO subsidiaries at the same time. It does not prove investment opportunities were scarce.
More telling, in our view, is LaSalle’s failure to ask ABN AMRO for additional capital. There is no evidence that LaSalle’s management sought or discussed seeking a capital infusion from ABN AMRO in order to pursue untapped investment opportunities. Nor is there evidence that management viewed the thrift’s capital position as constrained or that investment opportunities outstripped available resources. A fair inference is that management did not seek additional capital because the thrift maintained a sufficient amount to pursue all available and desirable investment opportunities during the relevant period.
Defendant also argues that plaintiff’s actual leverage ratio, which exceeded the regulatory leverage ratio minimum during the foregone-asset period, proves that investment opportunities were scarce.19 LaSalle’s leverage ratio surpassed the 5% regulatory minimum by at least 73 basis points (0.73%) every quarter during foregone-asset period for an average quarterly excess of 137 points. In two different quarters, the ratio exceeded the minimum by more than 200 points. Plaintiff defends these figures by arguing that its excess capital served as a cushion against unforeseen vagaries inherent in the market. On previous occasions, we have recognized that prudent thrifts maintain capital cushions above regulatory leverage ratio minimums. See, e.g., Commercial Fed., 59 Fed.Cl. at 347; Home Sav., 57 Fed.Cl. at 721. Evidence of a leverage ratio greater than the regulatory minimum, alone, therefore, does not establish scarcity of investment opportunities. The question is whether plaintiffs leverage ratio exceeded a reasonable capital cushion.
Plaintiff had its own internal capital cushion policy during most of the foregone-asset period. This policy prescribed a cushion 50-to 100-points above the regulatory minimum. Plaintiffs leverage ratio surpassed this policy fourteen out of twenty quarters during the foregone-asset period, an average quarterly excess of 37 points.
Mr. Roberts explained that an effective leverage ratio is not fixed by a hard and fast capital cushion policy and that the thrift may exceed such a policy when appropriate. According to Mr. Roberts, plaintiffs actual leverage ratio was based on management’s subjective judgment, taking into account asset riskiness, liquidity, ratios maintained by members of the thrift’s peer group, and other factors. His exercise of judgment “always left [the thrift] well above its stipulated internal mínimums.” Tr. at 177. Furthermore, Mr. Barefoot Bankhead, an expert witness for defendant,20 admitted on cross-examina*104tion that a 7% leverage ratio falls within the range of ratios maintained by most prudent thrifts — plaintiffs average ratio during the foregone-asset period was 6.4%. We conclude that a leverage ratio exceeding internal capital cushion policy, alone, therefore, does not prove that investment opportunities were scarce throughout the foregone-asset period.
More probative, however, is the leverage ratio’s drop near the end of the foregone-asset period. Both LaSalle and ABN AMRO maintained policies requiring the thrift to lower its leverage ratio by returning capital to the parent whenever investment opportunities waned. Pursuant to these policies, the thrift returned capital by paying ABN AMRO a $27 million dividend in December 1996. The thrift’s board of directors approved the dividend on October 16, 1996. This dropped the thrift’s leverage ratio nearly 100 points.
According to Mr. Roberts, this dividend was declared because it was requested by ABN AMRO, not because LaSalle viewed investment opportunities as scarce. According to Prof. James, the leverage ratio’s large drop near the end of the foregone-asset period was a response to a change in risk and economic climate. These explanations, however, are at odds with LaSalle policy, which required a dividend payment when investment opportunities waned. Furthermore, there is no evidence that either risk or the economic climate had changed. The 1996 dividend, the first of its kind that constituted a return of capital, suggests that, at least two-and-a-half months prior to the close of the foregone-asset period, LaSalle could no longer profitably deploy all of its capital. We therefore find it likely that a scarcity of desirable investment opportunities prompted plaintiff to return capital to ABN AMRO.
Furthermore, it is not probable that the evaporation of investment opportunities was sudden. It appears more likely that Taiman would have lacked investment opportunities at an even earlier point during the foregone-asset period. This scarcity is problematic for the lost profits claim because the claim is balanced upon plaintiffs simple assumption that opportunities would have been numerous enough to support its earnings projection throughout the foregone-asset period. Plaintiff offers no evidence to substantiate the availability of allegedly foregone investment opportunities. Without such a showing, we are unable to conclude with any confidence that plaintiff could have leveraged its unmitigated supervisory goodwill at a 7% rate throughout the foregone-assets period.
The circumstantial evidence of limited attractive investment opportunities, while perhaps not fatal in itself, creates a degree of uncertainty in plaintiff’s optimistic model that is difficult to ignore in light of a lack of detail concerning investment opportunities that would have been pursued.21 Plaintiffs failure to identify any specific types of foregone assets, or the proportion in which they would have been held in the but-for world, undermines the reliability of the foregone-asset calculation.
The earnings-on-foregone-assets projection relies on another calculation as well — the rate at which foregone assets would have produced earnings. The ROAA is the average of LaSalle’s rate of return for every quarter in a particular year. According to plaintiff, the ROAA increased yearly until 1996, when it experienced a slight decline.
Defendant’s main challenge to the ROAA calculation is that it does not account for the principle of diminishing marginal returns. Prof. Anjan V. Thakor, Ph.D.,22 explained that the law of diminishing marginal returns is a firmly established economic principle. A business typically will earn a lower rate of return on the 101st dollar invested than it does on the first 100 dollars. This is because it pursues the most profitable opportunities first. The rate of return decreases as investments increase, assuming all other economic variables remain unchanged.
*105An examination of LaSalle’s performance during the foregone-asset period demonstrates the marginal rate of return principle. As discussed above, under the Taiman business strategy, mortgage-backed securities played an increased role in the thrift’s portfolio when core residential mortgage opportunities grew scarce. Furthermore, borrowed federal funds accounted for a greater proportion of the portfolio’s funding when consumer deposits declined. Prof. Thakor testified that, during the foregone-asset period, plaintiffs profits on these marginal assets and liabilities were not as high as the profits on its core assets and liabilities. For example, Mr. Bankhead testified that the yield on mortgage-backed securities during the early months of the foregone-asset period was lower than the yield on residential mortgages. According to the thrift’s March 1993 Operating Report, the yield on its mortgage portfolio exceeded the yield on its security portfolio by at least 0.71% every month between March 1992 and March 1993. Mr. Bankhead felt that this yield differential was consistent with his personal experience: “[M]ortgagebaeked securities generally tend to have lower yields than [residential mortgages] do.” Tr. at 1113. Plaintiffs documents also demonstrate that borrowed federal funds used to purchase mortgage-backed securities were more expensive than plaintiffs core funding option, consumer deposits. In his expert report, Prof. Thakor demonstrates that this marginal funding source cost the plaintiff more every year during the foregone-asset period. Def.’s Ex. 887 Attach. 7 (increasing from 0.26% in 1992 to 0.77% in 1996). In short, in plaintiffs actual portfolio during the foregone-asset period, the spread between liabilities and assets was smaller at the margin than it was for core liabilities and assets.
It is worth noting that, while the ROAA calculation incorporates LaSalle’s lower marginal rate of return — the ROAA for any year during the foregone-asset period is an average of both higher-return core assets and liabilities and lower-return marginal assets and liabilities — the ROAA methodology does not account for the likely increase in the proportion of marginal investments between portfolios of actual and foregone assets. It is also irrelevant that the ROAA is significantly lower than the return rate of other Illinois thrifts and slightly lower than the return rate of thrifts nationwide during the foregone-asset period. These rates were derived from returns on actual assets in the real world. They shed little light on the accuracy of plaintiffs projected rate of return in the but-for world. Finally, it is also no answer that the ROAA grew nearly every year of the foregone-asset period. The marginal return principle operates at every discrete point in time. Using the ROAA to project future rates of return over a period of time does not reflect this reality despite the ROAA’s continued growth throughout the lost-assets period.
Plaintiffs model does not account for the evidence discussed above, nor does its deal with Prof. Thakor’s testimony that growing the size of an asset base necessarily increases a thrift’s exposure to risk, even if a constant proportion of certain types of liabilities and assets is maintained. The varied yields, maturity dates, and funding sources of each liability and asset create potential mismatches on a thrift’s balance sheet. These mismatches carry inherent risks that affect a thrift’s calculus when considering the acquisition of future liabilities and assets. Therefore, it cannot be assumed, without some better proof, that a thrift can simply grow its portfolio by replicating a successful formula of liabilities and assets without incrementally increasing risk.
Once again, the absence of any detailed information about the particular assets foregone forces us to find it likely that the investment of additional capital would have been less profitable than the investment in actual assets proved to be.23 The fact that the thrift had already encountered diminished returns on marginal liabilities and assets in its actual portfolio indicates that any increase in portfolio size would have required a greater reliance on marginal investments. We are forced to presume that the overall mix of *106but-for assets would have yielded a lower rate of earnings than was realized on the actual asset mix held by the thrift during the foregone-assets period. Plaintiffs ROAA calculation thus obscures substantial uncertainty. We are unable to ascertain the degree of that uncertainty due to the lack of evidence of the types and proportion of assets foregone. We hold that the projected ROAA is not reasonably certain.
D. Conclusion
We conclude that the twin uncertainties of both the amount and composition of foregone assets, and the hypothetical rate of return preclude recovery under plaintiffs earnings-on-foregone-assets projection. We nevertheless award plaintiff a portion of the $13.5 million in damages identified in LaSalle I. The Federal Circuit directed this court to determine the size of the offset necessary to account for ABN AMRO’s mitigating $300 million capital infusion — it did not disturb our holding concerning wounded-bank damages, foregone-mortgage-servicing damages, or profits lost due to the shrink in assets. Plaintiff tailored its earnings-on-foregone-assets projection to the Federal Circuit’s holding by backing $300 million out of the supervisory goodwill that would have been leveraged in the but-for world. There is thus no basis to offset earnings on the $300 million against the amount of our previous findings. It is necessary only to eliminate $5.23 million, which represents the overlap between the period of the shrink and the foregone-asset period. Accordingly, we award plaintiff $5,008,700.00 for wounded-bank damages, $1.9 million for foregone mortgage-servicing earnings, and $1.38 million for profits lost on the shrink.
II. Cost-of-Replacement-Capital Damages
A. The Remand Order
Our earlier opinion rejected plaintiffs claim for damages associated with partial mitigation provided by ABN AMRO’s $300 million capital infusion. We agreed with plaintiff that “the cost of replacement capital can serve as a valid theory for measuring expectancy damages in the Winstar context” but held that there was insufficient evidence that the theoretical dividend model constructed by plaintiff reflected any binding obligation of the subsidiary thrift to pay the parent for the use of the capital. LaSalle I, 45 Fed.Cl. at 103, 112. The Federal Circuit reversed, holding that:
[T]he court erred, for the cost of capital does not depend on whether payment is made as debt, or out of anticipated future earnings. An investor does not make a gift when the expected payment is dividends out of future earnings. All capital raised by a corporation has a cost, and it is well established that the payment of dividends is a capital cost.
LaSalle II, 317 F.3d at 1375 (citations omitted). The court went on, nevertheless, to reject the model offered at trial, in which the cost of dividends was represented by an assumed 12% rate, the return that ABN AMRO expected to earn on its investments in subsidiaries. According to the court, “it does not reflect the actual experience that the dividends were paid out of earnings, and that the earnings appear to have exceeded the hurdle rate as well as Talman’s projected earnings but for the breach.” Id.
In determining whether the thrift incurred a net cost of replacing supervisory goodwill, the Federal Circuit noted that “the benefits of [the] capital must be credited, as mitigation due to the replacement of goodwill with cash.” Id. Finally, the court also preserved an argument offered by the government that, because the dividends were paid to the corporate parent, there was no injury.
B. Foreseeability and Causation
We held after the first trial that “plaintiff has established the general foreseeability of these types of damages.” 45 Fed.Cl. at 89. We also held that “Taiman was forced by OTS to raise capital following the breach as a condition of approval of its capital plan____ Taiman actually did mitigate its damages by raising the necessary capital, by negotiating a cash infusion from ABN AMRO.” Id. at 103. The second trial does not prompt us to reconsider these general findings of foreseeability and causation as they relate to the cost-of-replacement-capital model. It was *107plainly foreseeable at the time of the assistance agreement that, in the event supervisory goodwill became worthless, Taiman would have to replace goodwill with tangible capital, at least to the extent that it propped up minimum capital ratios.
The remand order thus completes the causal connections. $300 million in capital had to be infused because of the breach. According to the Federal Circuit, “[a]ll capital has a cost.” LaSalle II, 317 F.3d at 1375. In this case that cost took the form of dividends. Some portion of post-breach dividends, therefore, must be attributed to the $300 million capital infusion.
Defendant has also made other arguments which amount, in substance, to challenges to particular aspects of causation. We examine them below in the context of determining whether plaintiff has proven its damages with reasonable certainty.
C. The Cost-of-Replacement-Capital Model
The rationale behind plaintiffs cost-of-replacement-capital model and its general outline are as follows: The principal difference between the actual thrift and the thrift in the but-for world was the actual thrift’s replacement of cost-free supervisory goodwill with a costly $300 million capital infusion. LaSalle pays for the infusion’s cost via dividends distributed to ABN AMRO. In light of the remand, some of LaSalle’s actual and projected dividend costs must be associated with the infusion. Dividends, however, are a recoverable cost only to the extent that they can be attributed to the initial $300 million infusion, not to one of ABN AMRO’s later investments in the thrift. Furthermore, while the infused capital had a cost, it also generated two types of return. First, it permitted the thrift to support a portfolio of assets and liabilities. In that respect it duplicates the function of supervisory goodwill. There is no need to account for the return on that portfolio or the dividends associated with it. In addition, however, the cash infusion, as opposed to its supervisory goodwill corollary, is an income-generating asset. Plaintiffs model must account for the earnings on this asset.
Plaintiff assumes that there should be no theoretical difference between supervisory goodwill and tangible capital with respect to their ability to support a portfolio of assets and liabilities. In substance, we agree. There is no reason to assume that the portfolio supported by $300 million in tangible capital would not be substantially identical to a portfolio supported by an equal amount of supervisory goodwill in the but-for world. In this respect, the earnings generated by the capital infusion are therefore a wash. This is reflected in plaintiffs lost profits claim, which only seeks profits lost on the supervisory goodwill not replaced. The only difference between the actual and but-for thrifts, then, are the dividends prompted by the cash investment and the fact that the cash itself generates a return, a phenomenon which plaintiff accounts for in its model.
Plaintiff begins with the fact that, although Taiman would have had a portfolio virtually equal to the actual portfolio in the but-for world, it would have been leveraged by supervisory goodwill, an asset to which no dividends could be attributed. But-for dividends, in other words, would have been paid to then-existing shareholders who could not be said to earn a return on anything other than their stock investment. Actual dividends, in contrast, must be partially allocated to ABN AMRO’s capital investment in the thrift.
Even though costs are a product of the breach only to the extent they are attributable to the initial $300 million infusion, plaintiffs model attributes all post-breach dividends to the $300 million infusion and other, later, capital infusions. In Prof. James’ model, therefore, all dividends are allocated across all post-breach infusions of capital.
We agree with Prof. James’ model in part. To the extent dividends were paid subsequent to the $300 million capital infusion, the infusion was, of necessity, linked to some or all of those dividends. In that respect, we believe Prof. James’ position to be consistent with the essential holding of the Federal Circuit’s opinion. The but-for thrift would have held an asset — supervisory goodwill— that could not generate an independent ex*108pectation of a return, despite the fact that it would have permitted the maintenance of a substantial, income-producing portfolio. After the breach and the infusion of new capital, all of the dividends went to the new shareholder, ABN AMRO, which was expecting a return, not only on its stock investment of $97 million, but also on its $300 million capital infusion.
This can best be illustrated by ignoring, for a moment, the fact that the thrift’s ownership changed as a result of the capital infusion. If the previous owners, Talman’s shareholders, had infused $300 million in cash following the disallowance of supervisory goodwill, and if the thrift had been able to maintain the original portfolio without interruption, then the only difference between the thrift’s actual and but-for circumstances would stand out in starker relief: the thrift would be generating the same return on its original portfolio,24 but the investors would expect an additional return for the lost opportunity cost associated with tying up $300 million in cash. That money could have been placed elsewhere — instead it would be tied up in Taiman purely because of the breach.
The calculus should thus begin with the dividends actually paid after the takeover by ABN AMRO, and an attempt must then be made, as can best be reconstructed, to eliminate dividends not associated with the $300 million infusion. It is appropriate, therefore, for plaintiff to construct a model that attempts to isolate those dividends that are linked solely to the initial capital infusion. Whether plaintiffs model incorporates this adjustment, however, remains to be seen.
On retrial, plaintiffs cost-of-replacement-capital model is no longer hypothetical. Prof. James’ model is based on actual dividends paid between 1992 and the present, supplemented by dividends projected between the present and 2012, the last year supervisory goodwill would have been held. LaSalle paid, on a pre-tax basis, a total of approximately $1.774 billion in dividends between March 1992, the date of the capital infusion, and June 2003, when Prof. James made his calculations. From July 2003 to February 2012, when supervisory goodwill would have been finally extinguished, he projected future dividends with a present, pretax value of $2.9 billion.
Against this figure, Prof. James then applied a ratio consisting of the portion of ABN AMRO’s total investment in LaSalle that was represented by its initial $300 million infusion. The ratio reflects two phenomena. First, ABN AMRO invested more in the thrift than the initial infusion. In 1994, for example, an additional $134 million was contributed to facilitate LaSalle’s acquisition of branches from Home Savings. In addition, in 1995, LaSalle merged with Cragin, which had $449 million in equity. Prof. James concedes that the later infusions of additional cash came with their own, proportionate, expectations of a return; they effectively dilute the attribution of dividends to the cost of mitigating the effects of FIRREA. Second, he had to adjust the $300 million initial balance of ABN AMRO’s contribution to reflect the fact that the replaced goodwill would have declined over time. Applying these two factors, the ratio begins at 100% and ends in 2012 at 0.1%.
Another critical element in Prof. James’ analysis is a recognition that dividends that reflect a return of capital are not properly claimed as damages. LaSalle paid two types of cash dividends: mandatory and special. This characterization is based on the thrift’s internal policy and is typically reflected in the company minutes authorizing a particular payment. Prof. James drew no real significance from these characterizations. Instead, he viewed the retention of earnings sufficient to support a dividend as evidence that the dividend was a return on capital. If sufficient earnings were not retained, then any payment in excess would, perforce, be a return of capital. As we will see below, the parties disagree sharply about whether he properly categorized dividends as a return on capital versus a return of capital.
Prof. James concluded that the dividend costs associated with the $300 million infu*109sion were $423.5 million through June 2003 and that future costs, expressed at present value, would be $98 million. These figures, however, had to be adjusted to recognize the benefit attributable to the replacement of supervisory goodwill with cash. This is consistent with the remand direction of the Federal Circuit: “[T]he benefits of capitales cash value] must be credited, as mitigation due to the replacement of goodwill with cash.” 317 F.3d at 1375. Because cash is inherently fungible and is not traced "within the institution, Prof. James had no way of attributing a particular return to it. He therefore had to construct a hypothetical offset. He took the position that the most appropriate measure of return, for the period 1992 through 2003, was LaSalle’s actual return on assets. For the period after June 2003, he made an estimate based on the yield on Treasury securities. In sum, he supported a claim for cost-of-replacement-eapital damages of $203 million. This calculation reflects asserted damages on a pre-tax cost, in effect adjusting up for an assumed effective federal and state income tax rate of 40%.
D. Defendant’s Response
Defendant offers several challenges to Prof. James’ calculations.25 The most comprehensive is the criticism that Prof. James does not account for the profits earned on the portfolio leveraged by the capital infusion. In defendant’s view, “LaSalle seeks all dividend costs related to the $300 million— dividend costs attributable to leveraging the $300 million and dividend costs attributable to the direct investment of the $300 million— but only accounts for a portion of the earnings benefit of the $300 million — the benefit related to the direct investment of the $300 million — ignoring the earnings benefit from leveraging the $300 million.” Def.’s Post-Trial Reply Br. at 48. According to defendant, this fundamental “‘asymmetry’ is the crux of the problem with LaSalle’s model ... since the law clearly provides that a plaintiff is only entitled to its incremental mitigation costs less its incremental mitigation benefits.” Id. at 48-49 (citations omitted). Defendant is correct in its understanding of the model’s mechanics and its statement of the law, but it is incorrect in its conclusion that the former is inconsistent with the latter.
Prof. James does not have to account for earnings on that portion of plaintiff’s actual portfolio that was supported by the capital infusion because they equal those that would have flowed from the government’s original promise. Those earnings would have accrued to the but-for thrift because supervisory goodwill was capable of sustaining an identical portfolio. Requiring plaintiff to account for those earnings in its model would defeat the purpose of mitigation.
Defendant also contends that “LaSalle failed to prove that it paid a dime more in dividends as a result of the breach than it would have paid in the absence of the breach.” Def.’s Post-Trial Reply Br. at 43. Plaintiff responds that it does not matter whether LaSalle paid more in dividends before the breach than after. As Prof. James testified, even if Taiman would have paid, in the but-for world from 1992 through the present, more dividends to its prior shareholders than Taiman actually paid to ABN AMRO, plaintiffs would still claim the cost of dividends attributable to the $300 million infusion. This argument exposes a fundamental distinction, not only between the parties’ respective experts, but, we believe, between the government’s position and the Federal Circuit’s remand order.
Defendant’s argument is that all dividends that would have been paid to shareholders but for the breach must be treated as an avoided cost in the actual model. It therefore contends that causation is not established if the cost-of-replacement-capital mod*110el does not account for those dividends. Prof. Thakor points out, for example, that common stock dividends of $1.9 million were paid in 1988 and $2.4 million in 1989. By projecting the growth of these dividends into the future, defendant swamps plaintiffs damages claim with avoided costs.
Defendant’s argument fails to account for the fact that the thrift’s sole shareholder in the real world, ABN AMRO, invested more in the thrift than but-for shareholders had. Therefore, a portion of the actual dividends paid to ABN AMRO must also be attributable to this capital infusion. No matter how large but-for stock dividends would have been, they cannot be presumed to cancel out entirely the thrift’s actual dividend payments to ABN AMRO. If defendant’s argument is taken to its logical conclusion, plaintiff could never recover in a cost-of-replacement-capital model because all but-for earnings are assumed to be paid out as dividends and all but-for dividends are treated as a cost. To do so would obviously defeat the purpose of the remand.
We view defendant’s argument as a reprise, in a different form, of a point made by Prof. Merton Miller, Ph.D.,26 during the first trial. Over the entire life of a corporation, dividends can never exceed the sum of earnings and net investment. By definition, a thrift can only pay out in dividends what it earns as net income. He thus took the position that there could be no net costs flowing from a cost-of-replacement capital model, other than transaction expenses. There is, however, room for plaintiff’s claim within the bounds of this economic principle. As even Prof. Thakor conceded at the second trial, because plaintiff’s model is limited to a finite period of time, dividends can exceed earnings.
Moreover, defendant’s argument assumes that earnings are never retained to grow the institution. In the actual world, LaSalle’s opportunity to do so was reduced because the but-for thrift would not have had to pay the particular dividends that plaintiff is claiming, at least at the time they were paid. Whatever other dividends plaintiff might have paid to former shareholders if there had been no breach, they cannot be equated, in full and within a finite time period, to the actual dividends associated with the $300 million.
Defendant’s argument, in any event, is incompatible with the remand order. If all actual dividends are canceled out by avoided costs, there is no room to implement the plain direction of the Federal Circuit to isolate costs associated specifically with the new capital infusion. The remand opinion does not give us the option to ignore ABN AMRO’s expectation of earning an additional return on that investment. To the extent, therefore, that defendant’s argument eliminates any possible net damages, we reject it. We believe, however, that some accounting must be made of pre-breach dividends. Admittedly this is something plaintiffs model does not do. As we discuss below, plaintiffs model must therefore be adjusted.
Defendant also advances more particularized challenges to Prof. James’ method of calculating damages. With respect to the total amount of dividends paid on all capital after the breach, defendant contends that Prof. James’ figure is inflated. It points to the fact that ABN AMRO and its subsidiaries distinguish between mandatory and special dividends. According to defendant, LaSalle’s own characterization of a dividend as “special” means that it is something other than a return on ABN AMRO’s capital investment. Defendant asserts that virtually no costs were incurred because nearly all of the past dividend payments were treated as “special” on the company’s books. We disagree.
LaSalle is subject to ABN AMRO’s Capital Allocation and Dividend Policy. The policy’s apparent purpose is to provide discipline to the movement of funds between ABN AMRO and its subsidiary companies in order to generate the maximum return on capital. The policy characterizes “mandatory dividends” as quarterly payments “equal in sum to one-third of the subsidiary’s budgeted net income.” Mandatory dividends have to be paid *111regardless of the impact on capital ratios. Special dividends, on the other hand, are “essentially ‘excess’ capital, over and above that which meets the required level of 50 basis points above the well-capitalized minimums. Special dividends will be allowed only at the discretion of [ABN AMRO’s chief financial officer], and only if the demand and supply of capital equilibrate----”
We see no necessary connection between LaSalle’s characterization of dividends as special or mandatory and Prof. James’ calculation of the total amount of dividends paid. Either type of dividend is still a return on investment, unless, as Prof. James concedes, the investor’s capital is returned. The fact that capital may be “excess” in the eyes of a subsidiary does not have a necessary connection to what constitutes the parent’s initial investment. When determining whether a dividend is a return on capital or a return of capital, the relevant criteria is the source of the funds, not the name given to the dividend. So long as the money paid is not a return of capital, it has to be attributed, at least in part, to the parent’s capital investment. Prof. James concluded that a return of capital did not begin until December 1996, based on a careful review of LaSalle’s capital accounts.
We note that ABN AMRO maintained a 12% “hurdle rate,” which was the rate of return it expected to receive on all funds invested in its subsidiaries. It is undisputed that LaSalle never met its parent’s hurdle rate. In fact, the dividends claimed here amounted to less than an average annual return of 7%. This is circumstantial evidence that the initial $300 million infusion did not diminish over time due to the return of capital, at least until December 1996. We are mindful, as well, of the Federal Circuit’s rejection of any suggestion that only “mandatory” dividends were a relevant cost of capital. LaSalle II, 317 F.3d at 1375.
The government also offers the related argument that each dividend must be evaluated independently from the standpoint of causation. In other words, if the impetus for a particular dividend cannot be traced to the breach, then the plaintiff cannot show a causal link. It points, for example, to a September 1993 dividend of $93.1 million, which defendant contends was prompted by unique business considerations that had nothing to do with the breach. This dividend, described in a 1999 Report of Examination by OTS, was part of a complex series of transactions by which one of LaSalle’s business components merged with another ABN AMRO subsidiary.
In effect, defendant is asking for proof that the dividend would not have been declared when it was, but for the breach. Such proof, however, is not required. The breach need not prompt a particular decision to declare dividends. Whatever immediate business considerations spurred the decision to issue the $93.1 million dividend, we believe it sufficient that the breach caused the transfer of some portion of earnings, via dividends, to ABN AMRO because of the $300 million infusion.
Defendant finally argues that the quantum of dividends was manipulated to boost plaintiffs damages claim. There is no evidence to support this charge. We therefore accept Prof. James’ calculations with respect to the total amount of dividends constituting a return on investment — approximately $1.8 billion on a post-tax basis.27
As explained above, Prof. James makes use of the ratio between ABN AMRO’s total investment in LaSalle and its $300 million infusion. He applied this ratio to the total amount of dividends paid. Defendant contends that the ratio’s denominator, which represents ABN AMRO’s total investment in LaSalle, should include the $97 million ABN AMRO paid to acquire Talman’s outstanding stock. Plaintiff responds that, from the thrift’s perspective, that cash was not available to earn a return and should not, therefore, inflate the ratio’s denominator. As Mr. Heitmann testified, the money paid to Talman’s shareholders was never part of the thrift’s capital account. Unlike the $300 million, it did not earn LaSalle a return.
From ABN AMRO’s perspective, the anticipated return on its initial $397 million in*112vestment in the thrift — the $97 million stock purchase and $300 million capital infusion— could come from two sources: appreciation of the stock’s value and dividends. According to plaintiff, defendant’s argument is thus off the mark because ABN AMRO could have sought a return on its $97 million from the marketplace. When asked whether he expected a return on the $97 million, Mr. Heitmann’s answer was clear: “Absolutely.... [b]ut we had to get that in a different way than we got [a return] on the $300 million. We had to get that through the appreciation in the value of the stock [of the parent company, ABN AMRO] ____” Tr. at 499. He further testified that ABN AMRO’s North American operations, of which LaSalle was a major part, increased the parent’s market valuation during the relevant period.
Defendant, in turn, responds that the fact that the $97 million was paid to prior shareholders and was not held by the thrift as a working asset is irrelevant. ABN AMRO, by paying for the stock, acquired the thrift’s existing assets, on which it expected to earn a return. Furthermore, as defendant points out in connection with its cost-offset argument, Taiman paid dividends to its shareholders in 1988 and 1989, prior to the $300 million infusion. Plainly those dividends are attributable to the shareholder’s expectation that they would earn a return on their stock investment. The evidence at trial merely confirms that ABN AMRO expected a return on its stock purchase, independent of the separate $300 million infusion, and that it hoped to get at least part of this return through stock appreciation. There is no further evidence, however, as to the return ABN AMRO actually realized on its stock purchase through appreciation.
The question of how to account for the $97 million ABN AMRO spent acquiring Talman’s stock brings us back to our prior discussion concerning whether but-for dividends are avoided costs that offset plaintiffs claim. We are left with the following: the thrift, pre-breach, paid dividends without the capital infusion and presumably would have continued to pay dividends in the but-for world. Those dividends must be attributable to the thrift’s value prior to the breach, which was unaffected by ABN AMRO’s later investment. The shareholders simply changed names. The thrift’s pre-breach bricks, mortar, regulatory licensing, and portfolio of assets and liabilities were still available to bank managers to generate income after the breach. The infusion of $300 million, as well as subsequent infusions, merely represents a new and proportionate call on the dividends paid after the breach.
Thus, while we reject defendant’s argument that the but-for dividends are co-extensive with actual dividends, we agree that some portion of actual dividends should be attributed to the thrift’s ability to generate earnings independent of the ABN AMRO acquisition. While some of that return may have come in the form of stock appreciation, in the absence of any better evidence we treat the proportion of ABN AMRO’s total investment that is represented by the $97 million stock purchase as the best estimate of that return. Prof. James’ model, therefore, should be adjusted to increase the capital on which ABN AMRO earned dividends by that amount.
Defendant also contends that Prof. James’ model fails to account for the fact that replacing supervisory goodwill with cash confers a benefit on LaSalle beyond the earnings on cash. According to Prof. Thakor, an institution with cash as capital is viewed as a safer borrower than an institution that relies on supervisory goodwill. The result, he reckons, should be a savings in borrowing costs to LaSalle after mitigation. Prof. Thakor’s observation turns out to be purely theoretical. It is not based on any demonstrated actual experience of plaintiff or of similar banks. We rejected a similar argument in Home Savings for lack of proof. 57 Fed.Cl. at 729. The assertion receives no better support here. For defendant to succeed on its argument, it would have had to produce some quantitative proof of a comparative difference.
E. The Offsetting Earnings on Cash
The Federal Circuit’s remand opinion directs us to offset plaintiffs costs with the earnings generated by the cash infusion’s tangible character, as distinct from the intan*113gible character of supervisory goodwill. Prof. James’ model uses LaSalle’s actual rate of return on its assets from 1992 through June 2003 to calculate past offsetting earnings. Beyond that date, he projected future offsetting earnings equal to the yield on Treasury securities. For the first quarter of 1997, for example, LaSalle generated a 1.79% return. Multiplying the remaining mitigation capital, $265.5 million, by that rate resulted in a $4.8 million offset. Cumulatively, the model calculates earnings, both past and projected, of $220.5 million. Prof. James then uses this figure as an offset against the $423.5 million in claimed dividend costs, producing a net cost-of-replaeement-capital claim of $203 million for the period from 1992 through 2012.
In response, Prof. Thakor contends that past earnings, which totaled $199 million, should have been compounded, thereby substantially increasing their amount. He argues that, to the extent earnings were not paid out as dividends, they could have been reinvested. While reinvestment is always a theoretic possibility, plaintiff points out that the dividend costs attributable to the infusion were higher than the earnings solely on the cash itself. There is no basis, therefore, for assuming a hypothetical compounding of the income generated by cash.
Defendant also argues that future earnings should be calculated in the same manner as past earnings. Prof. Thakor testified that the thrift’s return rate on its portfolio over the previous fourteen quarters should be utilized, particularly because plaintiffs future dividend costs are similarly a projection from past experience.
In Prof. James’ report, he anticipates this criticism:
Using Treasury rates to estimate the cash benefits of the mitigation capital has the important feature of mimicking the risk characteristics of supervisory goodwill because both supervisory goodwill and Treasury securities are government promises.
I assume the mitigation capital earns a risk-free return based on Treasury securities with maturities consistent with the planned amortization of the supervisory goodwill.
Pl.’s Ex. 992, at 23. There is thus an inconsistency within Prof. James’ approach. In effect, plaintiff concedes that, until 2003, the difference between the dividends it paid and its actual return on assets has been a valid measure of the premium associated with substituting real capital for supervisory goodwill. After June 2003, however, it switches from the actual rate of return on its portfolio to a substantially lower, risk-free return rate to predict earnings into the future. The inconsistency is made sharper by the fact that Prof. James used the company’s actual experience to project dividends into the future.
We note, moreover, that Prof. James’ figures for the Treasury rate do not closely match the most immediate preceding actual return. He presumes that the return for 2004, for example, will drop from the actual rate of approximately 4.5% in 2003 to a Treasury rate of approximately 2.25%. Prof. Thakor offers the figure of 5.8% as an average over the preceding fourteen quarters. This has the appeal of more complete symmetry, but we note that the return over fourteen quarters dropped on a consistent basis, starting at approximately 7% and dropping to approximately 4.5%. With respect to the future return on cash, therefore, we believe a more conservative but realistic figure is projected from actual earnings. We therefore assign a presumed rate, beginning in 2004 of 4.5%, with the rate thereafter consisting of Prof. James’ Treasury rate increased by 2%.28
F. The Implications of ABN AMRO’s Control Over LaSalle’s Dividend Policy
The Federal Circuit preserved the question of whether it is relevant that the dividends at issue were paid to a single corporate parent. Defendant suggests that an “award of damages to the corporate parent would constitute a windfall.” Def.’s Post-Trial Reply Br. at 61. The contention takes up little of defendant’s briefing, and the issue was not, at least to outward appearances, the *114direct subject of any testimony or other evidence.
It is not entirely clear to the court if defendant contends that no subsidiary should ever recover when dividends are paid to a single controlling shareholder. Defendant disavows such an argument, but it nevertheless asks “whether a subsidiary should recover damages amounts paid to its parent corporation, when, based upon uncontradicted evidence presented at trial, the parent can exercise complete control to commandeer the very award that was intended to compensate for prior payments to the parent.” Def.’s Post-Trial Reply Br. at 62. Because defendant’s suggested answer to this question is “no,” it is not clear what defendant’s disavowal means. Every parent corporation can exercise “complete control” over a wholly-owned subsidiary, including over its dividend policy, as can a majority of diverse shareholders.
To the extent defendant offers any further explanation, it cites American Silicon Technologies v. United States for the proposition that “fungible financial assets invite manipulation.” 334 F.3d 1033, 1037 (Fed.Cir.2003). Defendant contends that it is a sufficient comprehensive defense here that “ABN AMRO always has controlled LaSalle’s dividend policy” and has used that control “to transfer capital freely among [its] subsidiaries ____” Def.’s Post-Trial Reply Br. at 61. The argument is a non sequitur. Either the parent-sub relationship is a per se defense, or it is not. We think it is not. There were no grounds offered here to “pierce” the corporate veil on the theory that LaSalle is not an independent entity or that dividends were manipulated in anticipation of litigation. It adds nothing to observe that the corporate family moved capital around to maximize collective advantage.
Absent some compelling evidence that the parent was generating a damages claim by contriving dividend payments, we reject the implications of defendant’s argument. There is no such evidence. The subsidiary was an independent corporate entity and is entitled to assert its claim.
G. The Tax Implications of a Cost-ofReplacemenl^Capital Award
Plaintiff assumes that the award will be subject to income tax. See Centex Corp. v. United States, 55 Fed.Cl. 381, 389 (2003) (“The need for gross-up assumes ... that any recovery here is taxable.”). In order to put plaintiff in the same position that it would be in if there had been no breach, plaintiff therefore claims that the damages should either be calculated on a pre-tax basis or should be “grossed up” to account for future taxation. Both approaches are consistent with our holding in Home Savings, a ease in which damages were awarded based on the cost of replacement capital and in which the award was adjusted on the assumption that it would be taxable. 57 Fed. Cl. at 730. Defendant, however, argues that an award will not be subject to income tax and should therefore not be adjusted.
Dividends were paid from net earnings after taxes.29 We take it from the government’s argument that it agrees that plaintiff would not be made whole if the award was not adjusted up for taxes but was subsequently taxed. Compensating plaintiff with only the net amount of its dividend costs would not make it whole if that award would then be subject to tax. The question is thus whether the award would be taxed. If we conclude that it would, then the appropriate adjustment must be made. Regardless, plaintiffs recovery should be approximately the same in the long run. Clearly, if we make the adjustment, plaintiff would be es-topped from disputing the taxability of the award.
Defendant makes the argument that the award would not be taxed because it is a return of capital. It cites the testimony of Dr. Grant M. Clowery, Ph.D.30 Although not venturing a legal opinion on whether an *115award here would be taxed, his knowledge of tax accounting matters is obviously relevant. Dr. Clowery began with the observation that the payment of a dividend has no tax consequences for the payor company because dividends are not a deduction from income. He went on to state that “a dividend is not an income item for a payor company, and so therefore an award that replaces dividends, which would be classified based on the facts and circumstances of the award, for accounting purposes, would not be included in income, and would therefore not be taxed.” Tr. at 1188-89. He also testified that “[a] dividend is a capital item, and in order to meet the requirements of [26 U.S.C. § 61] and be in net income, the item would have to be reflected in the income statement of the company. And since dividends are not ... so reflected, it is clear that it’s not an income item.” Tr. at 1191. Presumably Dr. Clowery means that an item that was not treated as a deduction from income in the year paid would not be treated as income in the year returned.
Dr. Clowery agreed with the statement that, “to the extent [a dividend was] a return of capital, ... it would not be taxable; it [was] not a taxable event.” Tr. at 1193. He also agreed that the relevant inquiry is “whether what [LaSalle was] paying out is capital.” Id. In his view, it was. It is clear, therefore, that Dr. Clowery did not differentiate between initial or paid-in capital, on the one hand, and retained earnings. The return of any dividend would be not taxable per se, even one paid out of current-year earnings. In effect, Dr. Clowery proposes that the damages claim be treated as an asset lost and then restored. In his view, there would be tax consequences only if the company’s basis in the asset is less than the award amount.
In its post-trial brief, defendant expands on this argument. Defendant argues that “[a]ny recovery intended to replace capital generally is not considered taxable income.” Def.’s Post-Trial Br. at 108 (citing Milenbach v. Comm’n, 318 F.3d 924, 933 (9th Cir.2003); Raytheon Prod. Corp. v. Comm’n, 144 F.2d 110, 113-14 (1st Cir.1944)). In Raytheon, the court held that “[t]he test is not whether the action was one in tort or contract but rather the question to be asked is ‘In lieu of what were the damages awarded?”’ 144 F.2d at 113 (citing Farmers’ & Merchants’ Bank v. Comm’n, 59 F.2d 912 (6th Cir.1932)). The court went on to hold that “[w]here the suit is not to recover lost profits but is for injury to good will, the recovery represents a return of capital and, with certain limitations to be set forth below, is not taxable.” Id.
Defendant also cites the testimony of plaintiffs witness, Mr. Martin Eisenberg.31 Mr. Eisenberg was offered primarily to testify as to how plaintiff and its affiliated entities pay taxes and to support the mechanics of Prof. James’ adjustment for future taxes. Defendant objected to Mr. Eisenberg offering any opinion evidence as to whether the award would be taxable. We sustained that objection, but defendant then elicited the following testimony from Mr. Eisenberg:
Q. Before any dividend is paid by LaSalle Bank to ABN AMRO, the money used to pay that dividend is capital to LaSalle Bank; correct?
A. You mean a dividend is paid out of retained earnings?
Ct: All right. She’s asking whether that means that it would be identical to saying they’re paid out of capital?
A. From an accounting perspective?
Q. Is that a yes?
A. No, I’m asking you to clarify.
Q. I’m asking you, is it capital before it is paid out to LaSalle Bank?
A. Yes, it is a part of the capital, regulatory capital of LaSalle Bank before it’s paid out.
Tr. at 664-65. Defendant makes much out of his answer, “[y]es, it is a part of the capital,” linking that statement with the cases, discussed above, which suggest that returns of capital are not taxable. We believe defen*116dant improperly applies Mr. Eisenberg’s testimony.
First, Mr. Eisenberg’s statement is a small excerpt from an exceedingly confusing exchange which arguably could be excluded as beyond the scope of direct and beyond Mr. Eisenberg’s role as a fact witness.32 The court is persuaded that Mr. Eisenberg did not understood what he was being asked, and that his answers were theoretical. Defendant also ignores Mr. Eisenberg’s qualification to the cited testimony, in which he tried to explain that retained earnings might only be capital for regulatory purposes. He was not asked if LaSalle had treated the earnings from which any particular dividends were paid as an addition to the thrift’s capital.
In the final analysis, even if Mr. Eisenberg’s testimony is accepted at face value, we believe that defendant’s mechanical and simplistic use of it to invoke Raytheon and similar cases is off the mark. The gist of those decisions is that an award that replaces a capital asset should not be taxed. Such compensation would not be in the nature of a replacement of income — it would simply make the plaintiff whole. Cf. 26 U.S.C. § 104 (2000) (casualty losses and reimbursements for injuries not taxable as income). Here, plaintiff seeks the cost of replacement capital as part of a claim for expectancy damages. The dividend costs were an expense incurred in order to put plaintiff back into a pre-breach position with respect to its earning capacity. Plaintiff was, in effect, trying to protect its profits. We have no reason to believe that the Internal Revenue Service would treat the reimbursement of this cost item as a replacement of a capital asset.33
Defendant’s argument is, in any event, at odds with our holding above that plaintiff is entitled to reimbursement for dividend costs only to the extent that they were not a return of capital. By definition, therefore, the award here is not intended to make up for a loss of capital. Indeed, even defendant’s own arguments on the issue of whether a portion of the dividends were a return of capital assumed that some dividends constituted a return on capital. The latter were paid out of current or retained earnings which, for those purposes, defendant did not assert were paid out of capital.
In sum, there is no question that it would be unjust both to tax plaintiffs recovery and not adjust the award. It is only a possibility, and not a high one in our view, that the award will not be taxed. We cannot ignore the fact that, as a general proposition, amounts received as damages in litigation are taxable as income. See, e.g., United States v. Burke, 504 U.S. 229, 233, 112 S.Ct. 1867, 119 L.Ed.2d 34 (1992); Jones v. Corbyn, 186 F.2d 450, 453 (10th Cir.1950); Catron v. United States, 582 F.Supp. 8, 10 (N.D.Okla.1983). In addition, if the matter is contested, plaintiff would have the burden of proving an exemption. See Getty v. Comm’n, 913 F.2d 1486, 1492 (9th Cir.1990) (‘When contesting a deficiency determination by the Commissioner, the burden of persuasion rests on the taxpayer. This burden requires the taxpayer to show the merits of his claim by at least a preponderance of the evidence.” (citing Rockwell v. Comm’n, 512 F.2d 882, 885 (9th Cir. 1975))). We hold, therefore, that the award should be adjusted to account for subsequent taxation.
Defendant directed its main challenge of Prof. James’ tax adjustments to his use of a 40% effective tax rate. Defendant questions the validity of such an assumption and points to possible ways plaintiff or its parent might shelter an award from taxation in whole or part because LaSalle files a consolidated tax return with its affiliated and parent corporations. Defendant argues that the tax allocation agreement that binds LaSalle and these *117related companies permits the consolidated group to offset gains and losses to minimize taxes.
LaSalle is a separate entity, as discussed above. While the parent can use LaSalle’s gains or losses to optimal effect on the consolidated return, we view that fact as irrelevant. Even though the income LaSalle contributes to the corporate mix may find an offsetting loss elsewhere, the income’s taxable character, but for the unrelated loss, remains. It would unfairly penalize both LaSalle and the affiliated companies to anticipate factors unrelated to plaintiff that might impact the consolidated return. The net effect would be the same: the presence of income attributable to LaSalle causes the use or loss of tax write-offs elsewhere. Accord Centex Corp. v. United States, 395 F.3d 1283, 1291 (Fed.Cir.2005).
Prof. Thakor also questioned the way in which Prof. James went about grossing up the damages figure for taxes. Rather than use actual dividend costs and then leaving it up to the court to determine whether to make a tax adjustment, Prof. James divided dividend costs by 60% as a way to express costs in a pre-tax form. This assumes not only that the award will be taxed, but also that it will be subject to an effective tax rate of 40%. This 40% rate is the rounded composite of a presumed 35% federal income tax rate and a 4.5% state income tax rate.
Prof. Thakor is correct, in that Prof. James leaves no room for the court to adjust his 40% effective rate. If done according to Prof. Thakor’s method, the issues of taxability vel non, or the appropriate rate, could be dealt with once post-tax damages figures were determined. We would have preferred Prof. Thakor’s approach. If Prof. James is wrong at any point, including any element of damages, the calculation of the award must be returned to the parties. But the fact that Prof. James assembled his damages claim using pre-tax figures does not preclude a recovery, even if we disagree with his figures.
Defendant also argues that the 40% rate is purely speculative. This harkens to our earlier opinion, in which we declined to adjust the award for taxes because of a failure of proof. We found that plaintiff had not demonstrated that its marginal tax rate was 40% from 1993 through the present and that it would likely remain at that level into the future. LaSalle I, 45 Fed.Cl. at 110. Plaintiff has remedied that deficiency on retrial.
Mr. Eisenberg testified on this point. As tax director, Mr. Eisenberg is responsible for the preparation and filing of tax returns for both ABN AMRO and LaSalle. Even though the parent files a single consolidated return on behalf of its subsidiaries, each constituent company’s tax liability is separately calculated. By filing a consolidated return, the affiliated companies can use losses in one company to offset income elsewhere. At the corporate level, however, the Tax Allocation Agreement requires that each company recognize taxes as a separate company and remit taxes to the parent on that basis. In 2001, for example, LaSalle paid ABN AMRO $292 million to cover its federal and state income tax burden even though the parent avoided paying regular income taxes because of large losses elsewhere.
As a large corporation, plaintiff has been subject to continuous IRS audit since 1990. In addition, its intra-company tax allocation has been subject to scrutiny, but no apparent criticism, by bank regulatory officials. Moreover, the thrift has earned a positive net income every year since 1993. Plaintiff’s marginal federal tax rate, i.e., the rate at which incremental additional income is taxed, is 35%. This is because its net income has consistently exceeded the minimum cutoff of $18 million dollars. That rate has applied since at least 1993 and applies at the present time.
Plaintiff has also paid state income tax in Illinois, where it is headquartered, every year since 1993. Plaintiff or its own subsidiaries also pay taxes in several other states. Plaintiff uses a “rule of thumb” marginal rate of state tax of 4.5%. This is because most of its income is subject to the Illinois tax rate of 7.3%, but the net tax effect after deduction of these taxes on the federal return is 4.5%. The net tax effect of other state taxes is the same.
*118With respect to 2004, LaSalle, having shown a net income for the first quarter of $126 million, anticipates paying the 39.5% marginal rate. Barring a catastrophic event unrelated to the thrift, Mr. Eisenberg was aware of nothing in the short-term future that would make the company not profitable or that would result in a lower marginal rate. We have no basis to question his assessment.
Defendant points to the fact that, for two of the past three tax years, ABN AMRO paid the Alternative Minimum Tax (“AMT”). As Mr. Eisenberg explained at trial, however, the fact that the AMT “marginar’ rate is 20% does not affect the figures upon which plaintiff relies. The AMT is a parallel and alternative calculation. The taxpayer pays the higher of the AMT or normal income taxes. A marginal alternative minimum tax of 20% will of necessity produce a higher overall tax amount because it is calculated on a different base The only circumstance in which the 20% rate would become the actual effective rate is if there are many years of continual losses. Given LaSalle’s record, it is sufficiently unlikely that the rate can be ignored.
Finally, defendant makes much of the fact that ABN AMRO, LaSalle’s parent, paid no income tax in 2001. As we explain above, however, this is not relevant. If the court accepted the government’s argument, defendant would have succeeded in shifting part of its liability to the parent company.
We conclude, therefore, that Prof. James was justified in doing his calculations on a pre-tax basis, which has the same effect as grossing up a post-tax calculation with regard to the cost-of-replaeement-capital claim. We accept Prof. James’ calculation of dividend costs. The two adjustments identified above — inclusion of the $97 million stock purchase in the investment base and the higher anticipated return on cash — -prevent the court from calculating a final award on its own. The parties will be directed, therefore, to make the necessary calculations and present the court with the correct figure.34
CONCLUSION
Pursuant to the Federal Circuit remand, this court’s judgment of September 30, 1999, is vacated. We have examined the quantum of plaintiffs recovery under both the lost profits and cost-of-replacement-eapital models. Although the framework of plaintiffs earnings-on-foregone-assets projection was sound, it does not afford a reasonably certain measure of lost profits. We therefore reject that claim. The portion of our findings from LaSalle I that did not overlap the earnings-on-foregone-assets projection, $8,288,700, remains undisturbed. Therefore, we award that sum as damages for earnings lost on foregone mortgage servicing, profits lost during the shrink period, and wounded-bank damages. We also accept plaintiffs claim with respect to the cost of replacement capital, as modified above. The parties are directed to confer in an effort to stipulate to the correct amount of recovery.35 Whether or not successful, they shall file a joint statement on or before March 18, 2005, either indicating agreement on the correct amount of judgment, or proposing further steps to resolve any remaining issues.
. United. States v. Winstar Corp., 518 U.S. 839, 116 S.Ct. 2432, 135 L.Ed.2d 964 (1996).
. See LaSalle I for a complete discussion of the various mergers and acquisitions entered into by Taiman in the pre-FIRREA time period, and by LaSalle after the ABN AMRO acquisition.
. An extensive discussion of background facts can be found in our earlier opinions. LaSalle I, 45 Fed.Cl. at 69-77; see also Cal. Fed. I, 39 Fed.Cl. at 765-66, 779.
. The initial amount of supervisory goodwill was $912.6 million. The reduction to $463.9 million was a result of amortization.
. Prof. James holds an M.B.A. in finance and a Ph.D. in economics. He is the William H. Dial/SunBank Eminent Scholar in Finance at the University of Florida.
. Plaintiff assumes that profits earned on assets supported by the substitute $300 million are the same as those that would have been earned in a similar portfolio sustained by supervisory goodwill.
. In effect, this means plaintiff projects every $7 of foregone supervisory goodwill would have supported $100 of assets throughout the foregone-asset period.
. ROAA for any particular year is based on the division of a year's net income by the average actual assets maintained over the course of that year.
. The mandate rule is not applicable to these narrow situations: (1) evidence at a subsequent trial is substantially different from that at the initial trial; (2) controlling authority has since made a contrary decision of applicable law; and (3) the initial decision was clearly erroneous and would work a manifest injustice. Mendenhall v. Barber-Greene Co., 26 F.3d 1573, 1582 (Fed.Cir. 1994). The present case fits none of these exceptions.
. A plaintiff need only prove that the suffered injury was foreseeable and “of an amount that is not beyond the bounds of reasonable prediction.” 11 Corbin, Corbin on Contracts § 1012 at 76 (Interim ed.2002).
. Our finding that federal regulators expected Taiman to attain profitability was based, in part, on a memo authored by Brent Beesley, FSLIC’s director. LaSalle I, 45 Fed.Cl. at 89. In it, Mr. Beesley notes his expectation that Taiman would attain viability thanks, in part, to the supervisory goodwill on Talman’s books. Mr. Beesley’s trial testimony reconfirmed this belief. Dr. Jack Guttentag, a defense witness, also testified that the assumptions built into the agreement between Taiman and FSLIC in February 1982 indicated that FSLIC believed Taiman had a reasonable chance of becoming a viable entity.
. According to Mr. Theodore Roberts, the thrift’s former chairman, Taiman had been a "healthy, conservative, profitable thrift .... And suddenly, overnight, we became a pariah, and ... [regulators told] us what we could and could not do .... And in general, preventing us from exploiting opportunities that we otherwise would have.” Tr. at 130-31.
. Mr. Tempest is ABN AMRO’s former CEO.
. According to Prof. James, this ratio is a conservative representation of LaSalle's actual leverage ratio during the foregone-asset period. During that time, the annual average of LaSalle’s quarterly core leverage ratio ranged between 5.78% and 6.76%. The higher the percentage, the lower the ratio of foregone assets to capital.
. Lost profits are calculated on an annual basis. The supervisory goodwill balance for any given year is adjusted downward to account for the initial ABN AMRO capital infusion and plaintiff’s tangible capital deficiency. The balance of adjusted supervisory goodwill is averaged with the preceding year’s balance to yield average adjusted supervisory goodwill. See discussion supra Part I.A.
. Mr. Heitmann is the thrift's former president and the current vice-chairman of LaSalle Bank Corp., the thrift’s holding company.
. ABN AMRO’s initial $300 million infusion was required to attain regulatory approval of the corporation’s acquisition of the thrift. ABN AMRO also provided LaSalle with the capital *103necessary to purchase Cragin and Home Savings in two separate infusions.
. In support of its argument, defendant also cites papers authored by Prof. James concluding that thrifts affiliated with parent holding companies typically enjoy freer access to capital than their unaffiliated counterparts. Prof. James pointed out on cross-examination, however, that capital infused by a parent likely will be more costly to the thrift than capital raised from consumer deposits.
. Throughout the foregone-asset period, thrifts had to maintain a 5% leverage ratio in order to qualify as a well-capitalized institution. Minimum leverage ratios are established pursuant to FIRREA and its implementing regulations. 12 U.S.C. § 1464(t)(2)(A) (2000); 12 C.F.R. §§ 567.2(a)(2), 567.5(a), 567.8 (1992). Well-capitalized thrifts are defined in the Federal Deposit Insurance Corporation Improvement Act of 1991 and its implementing regulations. 12 U.S.C. § 1831o(a), (c) (2000); 12 C.F.R. § 565.4(b)(1)(iii) (1993).
. Mr. Bankhead is a C.P.A. and a director at Navigant Consulting.
. As discussed above, other decisions have held that this type of evidence was necessary to prove a claim for lost profits. See, e.g., Columbia First, 60 Fed.Cl. at 115-17; Commercial Fed.., 59 Fed. Cl. at 349 n. 29; S. Nat'l, 57 Fed.Cl. at 306.
. Prof. Thakor holds a Ph.D. in finance. He is the John E. Simon Professor of Finance at Washington Universily.
. See the discussion supra concerning Columbia First, Commercial Federal, and Southern National, as well as note 21 and accompanying text.
. In addition to this return, plaintiff would earn whatever income is derived from holding cash as capital rather than supervisory goodwill.
. Defendant no longer advances the argument, made at the first trial, that damages for the cost of replacing capital are, of necessity, limited to flotation costs. In the first California Federal appellate decision, the Federal Circuit affirmed a decision of the trial court limiting such damages to flotation costs. Cal. Fed. II, 245 F.3d at 1350; see also Bank United, 80 Fed.Appx. at 672. But we view that holding as confined to the peculiar proof presented in that case and not a ruling that, as a matter of law or elemental finance principles, such damages can never exceed flotation costs. We look, in any event, to the appellate decision in this case, which seems to be a repudiation of such a per se argument.
. Prof. Miller, now deceased, was a recipient of the Nobel Memorial Prize in Economic Sciences in 1990.
. Prof. James’ figures are ultimately offered on a pre-tax basis of approximately $3 billion.
. The same rate of return would be used to give a current discounted value.
. Prof. Thakor testified that, from a finance standpoint, “if [LaSalle] lost anything, it was after-tax profits. If they suffered a net cost it was an after-tax cost.” Tr. at 1431.
. Dr. Clowery holds a J.D., an M.B.A., and a Ph.D. in business and is also an unlicensed CPA. He is employed as an independent contractor that provides financial statement analysis in a number of contexts. He has taught in the accounting field.
. Mr. Eisenberg is a LaSalle Bank Corp. Senior Vice President and the director of taxes for the thrift and ABN AMRO.
. Plaintiff objected on both grounds. We sustained the objections but did not strike the testimony during trial.
. Plaintiff, unfortunately, accepted defendant’s premise in arguing that it had no basis in supervisory goodwill and hence the award would be taxed. Defendant compounds the confusion in its reply by arguing that "LaSalle presented no evidence that it paid taxes on the regulatory capital that it placed on its books as a result of its contract with the Government.” Def.’s Post-Trial Br. at 108. Both observations are irrelevant and confusing. The award of costs in connection with obtaining new capital does not substitute for the capital itself.
. Prof. James rounded the effective tax rate up to 40%. We direct the parties to use a 39.5% effective tax rate because it has greater precision.
. Defendant’s agreement is, of course, predicated on the opinion as written. It waives no rights to challenge the findings or holdings leading to any stipulated amount.