Opinion by Judge FLETCHER; Partial Concurrence and Partial Dissent by Judge NORRIS.
FLETCHER, Circuit Judge:These three related appeals arise from the collapse of a commercial real estate development project financed by industrial development bonds. Bondholders Eaton Vance California Municipals Trust, Municipal Securities Trust High Income Series 10, and Municipal Securities Trust High Income Series 11 sued the developer and affiliated companies, the trustee, and various law firms involved in the *1482bond issue. One of the bondholders, Eaton Vance, also sued the underwriter. The bondholders settled with the developer and the law firms, and the district court entered summary judgment in favor of the underwriter and the trustee. The bondholders appeal these judgments.
Eaton Vance’s claims against Bear, Stearns & Co., Inc., the underwriter, are based on alleged misrepresentations in the offering documents and alleged breaches of fiduciary duty. We affirm the district court’s dismissal of all of Eaton Vance’s claims against Bear Stearns except the sixth claim, for breach of fiduciary duty — although as set forth below, we reverse only as to a narrow segment of that claim.
The bondholders’ five claims against the trustee, First Interstate Bank of California, are based primarily on the Indenture of Trust, the contract which specified the terms on which the bond proceeds could be disbursed to the developer. The district court concluded that one of those claims had been waived, and that the breaches alleged under other claims did not cause the bondholders’ injuries. We affirm in part and reverse in part.
The trustee filed a protective cross-appeal against two of the settling law firms, contending that the district court’s order granting the law firms’ motion for a good faith settlement determination was based on the mistaken assumption that the trustee’s opposition to the motion was mooted by the summary judgment in its favor. Since our partial reversal of the summary judgment in favor of the trustee means that the trustee’s opposition to the good faith settlement determination is not moot, and since it is far from clear that the district court’s ruling was based on the merits, rather than on an assumption of mootness, we remand.
I
Eaton Vance v. Bear Steams
Background
At the end of 1986, Bear Steams underwrote a $9.7 million issue of industrial development bonds, the purpose of which was to finance the Kress Project, a commercial real estate development project in Long Beach, California. Industrial development bonds, or IDB’s, are issued by a municipality; the bond proceeds are lent to a private developer, and the municipality ceases to be liable on the bonds. The advantages attaching to municipal bonds, however, remain: the bonds are exempt from the registration requirements of the Securities Act of 1933,15 U.S.C. § 77c(a)(2), and even from the private anti-fraud provisions of the 1933 Act, 15 U.S.C. § 171(2); see Richard Jennings & Harold Marsh, Securities Regulation at 294-95 (6th ed. 1987). Moreover, the interest paid to the holders of IDB’s is tax-exempt. Id.
This latter advantage was largely swept away by the Tax Reform Act of 1986, Pub.L. No. 99-514,100 Stat. 2085, and this is in part what led to the complexity of the case before us: all parties involved in the bond issue were eager to close the deal before January 1, 1987, when the new tax laws went into effect, and in order to do so they had to move quickly in the last weeks of December 1986. Ultimately they were forced to structure the transaction in two steps: an initial closing in December 1986, and a second closing in April 1987.
The other twist to this case is that Richard Deringer, one of the principals of the developer Kress Associates, diverted most of the bond proceeds intended for the project to other companies he controlled. He has since pled guilty to two counts of bank fraud.
The major players involved in the issuance of the bonds were the City of Long Beach; Kress Associates, whose partners were Richard Deringer and Paul Klapper; First Interstate Bank of California (FICAL), the bondholders’ trastee; Bear Stearns, underwriter for the bond issue; and three law firms: Jones, Hall, Hill & White, representing the City; Freytag & LaForce, representing Kress Associates; and Brobeck, Phleger & Harrison, representing Bear Stearns.
The significant documents were the Purchase Agreement, a contract between the City and Bear Stearns in which Bear Stearns agreed to purchase the bond issue for face value, less a reoffering discount; the Loan Agreement, a contract between the City and *1483Kress Associates which set forth the terms of the City’s loan of bond proceeds to Kress; the Company Note, in which Kress agreed to repay the City; the Indenture of Trust, a contract between the City and FICAL as trustee for the bondholders, which assigned the City’s rights under the Loan Agreement and Company Note to FICAL, and set forth the terms under which bond proceeds could be disbursed to Kress; the Guaranty of Operating Deficits by Kress Associates’ partners, which was to provide security for certain of Kress’s obligations under the Loan Agreement; and the Surety Bond, which provided additional security for the guaranty.
The bonds were issued on December 23, 1986; bond proceeds were to be held in escrow by FICAL until Kress Associates satisfied various outstanding conditions of the Indenture — obtaining a construction contract, a surety bond, and title to the development property. The Indenture provided that if the conditions were not met by April 15, 1987, the bonds would be subject to mandatory redemption, to take place on June 1,1987.
Between December 1986 and March 1987, Eaton Vance, an institutional investor, placed orders for and subsequently “settled” or finalized the purchase of $4.5 million in bonds, in four installments. At the time it placed its first order, the only offering document available was a Preliminary Official Statement (POS), prepared by Bear Stearns through its counsel, the Brobeck firm. Before the settlement date for the first and all subsequent orders, Bear Stearns mailed out the Official Statement (OS), which superseded the POS. It is not disputed that Eaton Vance received the OS. Eaton Vance’s portfolio manager, however, testified that nobody at Eaton Vance read the OS in connection with the first two or three purchases, and that he himself could not recall having read the OS before making any of the purchases. The bulk of Eaton Vance’s claims against Bear Stearns rest on alleged misrepresentations and omissions in the POS and the OS.
Eaton Vance made its final bond purchase in mid-March, 1987. As the April 15, 1987 escrow deadline approached, Bear Stearns learned that the conditions for closing escrow had not been met. According to Bear Stearns, a Bear Stearns director contacted Eaton Vance and two other institutional bondholders and got their permission to extend the deadline by one week. According to Eaton Vance, there is a dispute of fact on this issue. The escrow conditions were apparently fulfilled by April 22, 1987, and the bond proceeds became available for release from FICAL’s Construction Fund.
A year later, Eaton Vance learned that Kress Associates had made very little progress on the construction project, despite having drawn down nearly all of the bond proceeds. Eaton Vance contacted Bear Stearns and requested rescission of its purchase contract. That request was refused. In October 1988, Eaton Vance and the other bondholders replaced FICAL with a different trustee. Kress Associates was declared to be in default on its loan agreement and note, and the project was abandoned. On December 1, 1988, Kress Associates’ semiannual debt service obligations under the loan agreement became due. When the interest which was owing on the bonds (to be met by Kress Associates’ payment) was not paid, the bondholders made demand on the surety bond. The bondholders were paid $485,000 by the issuer of the surety bond; the bond then expired.
Two of the four misrepresentations Eaton Vance complains of concern the surety bond and the related security instrument provided by the Kress partners, the operating deficits guaranty. Eaton Vance complains (1) that the offering documents promised that the surety bond would remain in place until June 1990, while in fact it expired in December 1988; and (2) that the OS failed to inform bondholders that the guaranty had been altered so that it expired upon Kress Associates’ default. The two other alleged misrepresentations concern Deringer’s past performance and current capacity to complete the Kress Project: Eaton Vance argues that the POS and the OS failed to disclose that Der-inger had neither the money nor the honesty to get the job done.
This lawsuit began in November 1988, when Eaton Vance, together with the other major bondholders, sued Kress Associates, Deringer, Klapper, and several of Deringer’s *1484other companies in Shawmut Bank et al. v. Kress Associates et al. Later, the bondholders added FICAL and the three law firms as defendants, and most of the defendants cross-claimed against one another. In September 1989, Eaton Vance sued Bear Stearns in a different lawsuit, alleging violation of federal securities laws and eight pendent state law claims. The two actions were consolidated for pretrial purposes.
The district court treated Eaton Vance’s claims against Bear Stearns in two groups: six claims for damages (violation of federal and state securities laws, common law fraud and negligent misrepresentation, breach of fiduciary duty, and breach of warranty) and three claims for rescission (violation of state securities laws, breach of fiduciary duty, and mutual mistake). On April 17, 1992, the court entered summary judgment dismissing the damage claims. This judgment was based on findings of fact and conclusions of law entered earlier in the lawsuit, when the district court determined that the losses Eaton Vance complained of, based on decline in the value of the bonds, were not caused by any misrepresentations made by Bear Stearns.
On June 12, 1992, the district court granted summary judgment dismissing Eaton Vance’s three remaining' rescission claims. After first holding that rescission was an improper remedy, the court went on to discuss alternate grounds for dismissal. It held that Eaton Vance’s fifth claim, for misrepresentation under the California Corporations Code, was barred by the statute of limitations; that the seventh claim, for breach of fiduciary duty, failed because Bear Stearns did not owe Eaton Vance such a duty; and that the ninth claim, for mutual mistake, failed because no such mistake had been made. With respect to the fifth claim, the court also held that reliance on the POS was improper, and that Eaton Vance had shown no misrepresentations in the OS. While the district court made these latter rulings in the context of the fifth claim only, Eaton Vance correctly points out that the reasoning behind the rulings pertains to all of the claims in which misrepresentation is alleged, including, notably, the federal securities law claim.
Discussion
The district court had jurisdiction under 28 U.S.C. § 1331, based on plaintiffs federal securities law claim; it had ancillary jurisdiction over the related state law claims. Jurisdiction in this court is proper under 28 U.S.C. § 1291.
We review the district court’s two summary judgment rulings de novo, and we may affirm on any ground supported by the record. In re Apple Computer Sec. Litig., 886 F.2d 1109, 1112 (9th Cir.1989), cert. denied, 496 U.S. 943, 110 S.Ct. 3229, 110 L.Ed.2d 676 (1990). Summary judgment is appropriate if there is no genuine dispute of material fact and the moving party is entitled to judgment as a matter of law. Id.
A. Violation of Federal Securities Laws
The elements of a claim under § 10(b) of the Securities and Exchange Act, 15 U.S.C. § 78j(b), have often been repeated: a plaintiff must establish “ ‘(1) a misstatement or an omission (2) of material fact (3) made with scienter (4) on which the plaintiff relied (5) that proximately caused his injury.’ ” McGo-nigle v. Combs, 968 F.2d 810, 817 (9th Cir.), cert. dismissed, — U.S.-, 113 S.Ct. 399, 121 L.Ed.2d 325 (1992) (quoting Huddleston v. Herman & MacLean, 640 F.2d 534, 543 (5th Cir.1981), aff'd in part and rev’d in part on other grounds, 459 U.S. 375, 103 S.Ct. 683, 74 L.Ed.2d 548 (1983)).
The district court held that the OS1 did not contain any misrepresentations. The *1485district court’s judgment may be upheld not only if there were no misrepresentations, but also if the misrepresentations and omissions complained of were not material: even an admitted misrepresentation is actionable only if it is material. See Basic, Inc. v. Levinson, 485 U.S. 224, 288, 108 S.Ct. 978, 987, 99 L.Ed.2d 194 (1988) (“It is not enough that a statement is false or incomplete, if the misrepresented fact is otherwise insignificant”).
The test for materiality is well established. A fact is material if there is “a substantial likelihood that, under all the circumstances, the omitted fact would have assumed actual significance in the deliberations of the reasonable [bondholder.” TSC Industries, Inc. v. Northway, Inc., 426 U.S. 438, 449, 96 S.Ct. 2126, 2182, 48 L.Ed.2d 757 (1976). While the issue of materiality is fact-specific and “should ordinarily be left to the trier of fact,” summary judgment may nevertheless be granted “in appropriate eases.” Apple, 886 F.2d at 1113.
We discuss the allegedly misleading aspects of the OS seriatim.
1. “Deringer’s track record”
The Kress Project failed because Deringer diverted money earmarked for construction of this project to other companies he owned. Eaton Vance argues that Deringer had a history of doing this, and that it was materially misleading for Bear Stearns to omit that fact from its offering documents. Eaton Vance relies on the deposition testimony of Deringer’s former bookkeeper, who stated that records she had kept for other companies owned by Deringer revealed inter-company transfers.
Bear Stearns responds that there is no evidence that there was anything illegal or even improper about these transfers. Eaton Vance replies that be that as it may, investors should have been given the opportunity to decide for themselves whether or not they wanted to put money into the hands of a man who had a history of transferring funds among his various companies, and that the question of whether the omission was material should have been left to the jury.
Cases addressing the materiality of facts pertaining to a company’s criminal conduct reveal by way of contrast the weakness of Eaton Vance’s position. In Roeder v. Alpha Indus., Inc., 814 F.2d 22 (1st Cir.1987), the First Circuit held that the fact that a company has engaged in bribery is material, whether or not the criminal conduct has yet become the subject of investigation. The court explained that even “[ijllegal payments that are so small as to be relatively insignificant to the corporation’s bottom line can still have vast economic implications,” since they can endanger all of a corporation’s business if they are discovered. 814 F.2d at 26. But this logic does not extend to conduct which is not alleged to have been illegal: Eaton Vance does not argue that Kress Associates’ reputation or good standing would have been endangered had it become known that Deringer had made intercompany transfers.
Eaton Vance would open the floodgates here: if intercompany transfers are relevant, then it is hard to know where to draw the line, and how to determine what aspects of business history are not material. Here, the transfers complained of did not involve Kress Associates itself, but rather transfers between other companies owned by Deringer. See In re Convergent Technologies Sec. Litig., 948 F.2d 507, 516 (9th Cir.1991) (“A company need not detail every corporate event, current or prospective- The securities laws do not require management ‘to bury the shareholders in an avalanche of trivial information — a result that is hardly conducive to informed decisionmaking’ ”) (quoting TSC Industries, 426 U.S. at 448-49, 96 S.Ct. at 2132) (other internal quotation marks and citations omitted). Eaton Vance identifies no limiting principle, which suggests selection based on a choice of whatever hitherto undistinguished facets of business history may appear, retrospectively, to be linked in some way to the subsequent disaster.
But such morning-after methodology cannot supply guidance to those who must compose offering documents in the present tense. Because Eaton Vance has articulated no other criterion of relevance, its failure to allege *1486that the prior transactions were illegal is fatal.2
2. Funds needed to complete the project
In a section entitled “Estimated Sources and Uses of Funds,” the OS lists $9,700,000 in bond principal and $265,000 in company contributions, for a total source amount of $9,965,000. The OS also estimates $9,965,000 in total uses. Eaton Vance argues that these figures are misleading because they do not reveal that Kress Associates had been unable to obtain additional funds needed to complete the project.3
We do not see how the evidence Eaton Vance relies upon shows Kress Associates’ claimed inability to raise the extra money. Eaton Vance cites to a letter from James Madden, Deringer’s financial adviser, to Tom Dashiell, Deringer’s attorney. This letter states that Deringer, Klapper and others had been involved in a building partnership since 1981; it nowhere reveals, however, that they were unable to raise needed capital. Neither does Madden’s deposition testimony reveal any inability on the part of Deringer to raise the money; that testimony shows only that money was needed for the project. If Eaton Vance’s argument is that the OS was misleading because it did not reveal a track record of inability to raise capital, then the argument fails because no such record has been shown.
Eaton Vance might alternatively be understood to argue that the deception lay in some suggestion in the OS that the needed capital was already available, when in fact it was not. While this argument is more plausible, there is once again a fatal lack of evidence: Eaton Vance has not shown that Kress Associates lacked access to the needed capital,4 but simply that the money was not ultimately forthcoming. In short, we cannot see where the alleged misrepresentation lies.
3. Operating deficits guaranty
The OS states that
[a]s security for [Kress’] obligations under the Loan Agreement, Paul F. Klapper and Richard L. Deringer, the general partners of the Company, will personally provide a Guaranty of Operating Deficits ... guaranteeing the payment of debt service obligations accruing under the Company Note or the Loan Agreement until [the ratio of Project income to debt service obligations is 1.1 to 1].
In a subsequent description, the OS states that
*1487[t]he general partners of the Company will execute and deliver in favor of the Trustee a Guaranty of Operating Deficits (the “Guaranty”). The Guaranty secures payment of the Company Note by the Company in the event that operating revenues from the Project are insufficient. The Guaranty will be released at such time as the Debt Service Coverage Test has been met.
The misrepresentation Eaton Vance complains of concerns a handwritten change made on the guaranty on the day the document was executed. The typed form of the document states that
[a]t such time as the Debt Service Coverage Test has been fully satisfied and the Required Debt Service Reserve has been deposited in the Debt Service Reserve Fund, the Guaranty shall terminate and be returned to the Guarantors.
The handwritten insert adds, after the phrase “at such time as,” the clause “either (a) there has been a default on the Company Note and/or the Loan Agreement and an acceleration of the Company’s obligations under the Company Note, or (b).” This language has the effect of ensuring that the guaranty terminates upon default and acceleration. Accordingly, when Deringer’s diversions were discovered and the trustee declared a default in October 1988, the guaranty expired. Eaton Vance argues that the handwritten change materially altered and in fact defeated the whole purpose of the guaranty, and that the OS (which was mailed out after the guaranty was executed) was misleading in its failure to reflect that the change had been made.
Bear Stearns’ response is that the language inserted into the guaranty did not alter but merely clarified its terms; hence the failure of the OS to mention the change is not a material omission. We agree. The very title of an “operating deficits guaranty” reveals that the guaranty will expire upon default, since once a project is in default there are no longer any operations, and operating deficits cease to accrue.
We acknowledge that certain language in the OS describing the guaranty is broader than the title “operating deficits guaranty” would indicate. The OS states that the guaranty secures “payment of debt service obligations accruing under the Company Note or the Loan Agreement,” and then later states simply that the guaranty “secures payment of the Company Note.” Since the Company Note obligates Kress Associates to repay the entire amount of the interest and principal, these passages might be read as stating that the guaranty extends to the full scope of those obligations. Indeed, it appears that Deringer understood the guaranty in this way. Mark Holmstedt of Bear Stearns testified that Deringer’s “interpretation of the guarantee as it was written was that he was guaranteeing all obligations under the loan agreement and company note such that if he totaled them all up they would [amount to] principal plus interest, or approximately $20 million, round numbers.” In fact, it was because Deringer interpreted the guaranty in this way and-did not like it that he asked for the amendment and the amendment was made.5
Whatever Deringer’s understanding of the term operating deficits guaranty may have been, however, Eaton Vance’s own bond analyst, who was involved in the bond purchase, supplied a definition of that term very similar to the one Bear Steams now promotes: the analyst stated that an operating deficits guaranty is a guarantor’s “assurance[ ] that he w[ill] contribute towards the operation of the project sufficient to pay for project expenses and debt service.” This is inconsistent with the idea that the guaranty would also secure principal and interest upon default.
In light of the testimony of Eaton Vance’s bond analyst, and more particularly in light of its source, we conclude that the phrase *1488“guaranty of operating deficits” is in this case sufficient in itself to override any language in the OS which might have suggested that the guaranty would secure more than merely interim debt. Consequently, we conclude that the OS was not materially misleading in its failure to refer to the December 23 insertion.6
4. Surety bond
The OS states that “[a]s support for the Guaranty, [Kress Associates] has covenanted to provide a Surety Bond or Alternate Credit Facility ... in an amount at least equal to $970,000.” The OS explains that the surety bond is to be drawn upon “in the event that the general partners of the Company default in their obligations under the Guaranty,” and that the initial term of the bond will expire on June 80, 1990.
The “covenant” referred to in the OS appears in the Trust Indenture, which states, inter alia, that no bond proceeds are to be released from the Construction Fund created by the trustee until the trustee has received a surety bond in the amount of $970,000 securing Deringer’s and Klapper’s obligations under the guaranty. The Loan Agreement, incorporated by reference, states that the surety bond shall have an expiration date of June 30, 1990. The Trust Indenture also states that before Construction Fund money may be released, the trustee must receive written approval from the “Original Purchaser”—that is, Bear Steams—of the “form and substance” of the surety bond and the other documents, “which approval the Trustee may conclusively rely upon.”
A surety bond was executed by Deringer, Klapper, and the United Pacific Insurance Company on April 20, 1987, during the one-week extension of escrow. A termination clause states that the bond is to expire at the termination of the guaranty, or on December 2, 1988, whichever comes first. On the same day that the surety bond was executed, Mark Holmstedt of Bear Stearns sent a letter to FICAL approving the form and substance of the surety bond and other documents.
After the default in October 1988, Kress failed to make its December 1, 1988 debt service payment of $485,000. The bondholders made demand on the surety bond and were paid $485,000. The next day, the bond expired under the termination clause. Eaton Vance argues that the OS was misleading because it stated that the surety bond would be in force for another 18 months, until June 1990.
As Bear Stearns points out, and as the district court recognized, the surety bond— with its premature expiration date—was executed several months after the OS was mailed out. Thus the OS contained no representations which were untrue at the time it was disseminated. It stated only that Kress had “covenanted” to provide a surety bond with a given expiration date, which in fact it had. While a representation pertaining to a future occurrence may be actionable if the speaker is aware of undisclosed facts tending to undermine its accuracy, Apple, 886 F.2d at 1113, Eaton Vance does not argue that at the time the OS was written Bear Stearns knew that Kress would execute a surety bond with a different expiration date than that set forth in the OS—or that Bear Stearns would approve it.
In order to overcome this temporal obstacle, Eaton Vance argues that Bear Stearns had a continuing duty to disclose material facts to the bondholders. But while such a duty exists both under federal securities law, Ross v. A.H. Robins Co., Inc., 465 F.Supp. 904, 908 (S.D.N.Y.), rev’d on other grounds, 607 F.2d 545 (2d Cir.1979), and under common law misrepresentation, Koch v. Williams, 193 Cal.App.2d 537,14 Cal.Rptr. 429, 431 (1961), the duty is clearly limited to pre-transaction occurrences. After the transaction is completed, it would simply do no good to update information that has subsequently become misleading. See Ross, 465 F.Supp. at 909; Barnett v. Kirshner, 527 F.2d 781, 783 (2d Cir.1975). Here, since the *1489sale of the bond issue by the underwriter was completed long before the surety bond was executed, there was no continuing disclosure duty under federal securities law, nor any parallel state law duty.
This does not mean, however, that in the circumstances of this case Bear Stearns had no post-OS duty: we discuss below the argument that its post-OS conduct is the basis of liability for a breach of duties distinct from an underwriter’s normal obligations. But since neither Bear Stearns’ representations with respect to the surety bond, nor any of the other alleged misrepresentations or omissions provides the basis for a claim under federal securities law, we affirm the district court’s summary judgment dismissing Eaton Vance’s first claim against Bear Stearns. Because we do so on the ground that the alleged misrepresentations were not material, or were not misrepresentations at all, we do not reach the question of loss causation.
B. Other Misrepresentation-Based Claims
Our conclusion that Bear Stearns did not make any misrepresentation of material fact also disposes of Eaton Vance’s second claim (for fraud), third claim (for common-law misrepresentation), fourth and fifth claims (for violation of California Corporate Securities Law §§ 25400, 25401, 25500, and 25501), and eighth claim (for breach of warranty). See McCormick v. The Fund American Cos., 26 F.3d 869, 884 (9th Cir.1994) (analysis of materiality under federal securities laws also disposes of plaintiffs claims for violation of state securities laws, fraud and deceit, and negligent misrepresentation); Grigsby v. CMI Corp., 590 F.Supp. 826, 833 (N.D.Cal. 1984) (same), aff'd, 765 F.2d 1369 (1985); Lynch v. Cook, 148 Cal.App.3d 1072, 196 Cal.Rptr. 544, 550 (1983).
C. Breach of Fiduciary Duty
In its sixth claim, Eaton Vance seeks damages for “Breach of Fiduciary Duty with Respect to Disclosure”; in its seventh claim, Eaton Vance seeks rescission for “Breach of Fiduciary Duty with Respect to Release of Bond Proceeds from Construction Fund.” The gravamen of the sixth claim, like that of the federal securities law claim, is nondisclosure. The seventh claim is based not only on nondisclosure but also on actions Bear Stearns took with respect to the April 1987 close of escrow. In both claims, Eaton Vance contends that Bear Stearns assumed a fiduciary duty by virtue of these escrow-related activities.
Although at one point Eaton Vance premised Bear Stearns’ purported fiduciary duty in part on federal securities laws regulating underwriters, it has largely abandoned that position,7 stating that “Eaton Vance does not contend that Bear Stearns’ status as an underwriter creates a presumption of fiduciary duty,” and that “[t]he issue is whether a jury could find from all the evidence that Bear Steams knowingly undertook] to act on behalf of and for the benefit of another.” Reply Br. at 4-5 (internal quotation marks omitted).
Thus Eaton Vance’s contention is that Bear Stearns assumed a duty, largely by virtue of two actions it took in connection with the close of escrow. First, Eaton Vance points out that it was Bear Stearns’ role *1490under the Trust Indenture to approve several documents — including the surety bond— the securing of which was a condition of closing and of disbursement of Construction Fund money to Kress. Second, Eaton Vance argues that Bear Stearns undertook to act as its representative when it told FICAL that the bondholders had agreed to extend the deadline for closing escrow by one week— which had the effect of preventing mandatory redemption of the bonds. In addition, Eaton Vance points to a letter Bear Stearns sent to the California Debt Limit Allocation Committee, apparently in response to concerns the Committee had expressed about bondholder security. In the letter, Bear Stearns told the Committee that it “stands behind” the bonds it sells, and “reviews each issue to ensure its quality, keeping foremost in mind the security of the bondholder.”
The parties agree that we should be guided by Committee on Children’s Television, Inc. v. General Foods, 35 Cal.3d 197, 197 Cal.Rptr. 783, 799, 673 P.2d 660, 676 (1983), in which the California Supreme Court held that in order for a fiduciary obligation to exist, a person “must either knowingly undertake to act on behalf and for the benefit of another, or must enter into a relationship which imposes that undertaking as a matter of law.”
1. The surety bond transaction
The evidence reveals that Bear Stearns undertook to ensure that the surety bond, together with other documents, was as the parties had agreed. The Trust Indenture provides that before Construction Fund money could be disbursed, Bear Stearns had to furnish written approval of the “form and substance” of the surety bond. Bear Stearns was not, as it points out, a signatory to the Trust Indenture (a contract between the City and FICAL), although Bear Stearns was a party to the Purchase Agreement, a contract it entered into with the City roughly contemporaneously. We do not know why Bear
Stearns was designated in the Trust Indenture as the party responsible for looking after certain of the bondholders’ interests— whether it was because Bear Stearns had agreed to perform this service as a favor for the City, because Bear Stearns wished to provide a service for the bondholders which would be its customers, or for some other reason. What is important is that Bear Stearns agreed to and eventually did render the service. As a result, FICAL, the trustee, was relieved of any obligation to ensure that the bondholders were given the protections FICAL had agreed they would be given, since the Trust Indenture provided that FICAL could “conclusively rely upon” Bear Stearns’ approval of the documents. Under California law, it is well established that “[a] person not required to perform services for another may sometimes do so, and in such case is under a duty to exercise due care in performance.” 6 Bernard Witkin, Summary of California Law § 868 (9th Cir.1988) (collecting authorities). A similar principle has been recognized in the situation where, as here, a party renders professional assistance to a trustee. The party which “undertakes a relationship as adviser to a trustee ... in reality also assumes a relationship with the beneficiary akin to that between trustee and beneficiary.” Morales v. Field, 99 Cal. App.3d 307, 160 Cal.Rptr. 239, 244 (1979). That principle applies with special force here because, as a result of Bear Stearns’ undertaking, the bondholders were deprived of protections they would otherwise have been afforded by FICAL, which was entitled to rely on Bear Stearns’ approval, and did. See W. Page Keeton, Prosser and Keeton on Torts § 56 at 381 (5th ed. 1984) (duty may be imposed where defendant makes situation worse by depriving plaintiff of help from other sources). We conclude that by undertaking the task of ensuring that documents affording protection to the bondholders were satisfactory,8 Bear Stearns assumed a duty to Eaton Vance.
*1491It is crucial to understand the limits of that duty. Bear Stearns undertook to ensure that certain bondholder protections provided for in the Indenture would in fact be in place. Its duty cannot exceed that imposed by the Indenture itself. The duty which is imposed under an indenture of trust is not a traditional fiduciary duty. The Second Circuit has explained that
[a]n indenture trustee is not subject to the ordinary trustee’s duty of undivided loyalty. Unlike the ordinary trustee, who has historic common-law duties imposed beyond those in the trust agreement, an indenture trustee is more like a stakeholder whose duties and obligations are exclusively defined by the terms of the indenture agreement.
Meckel v. Continental Resources Co., 758 F.2d 811, 816 (2d Cir.1985); see also Elliott Assoc. v. J. Henry Schroder Bank & Trust Co., 838 F.2d 66, 71 (2d Cir.1988).9 If the indenture trustee does not have common law fiduciary duties, neither does the third party which undertakes to perform a part of the trustee’s role.
Because Bear Stearns’ duty is limited in this way, it does not include, as a corollary, the higher standard of disclosure which Eaton Yance seeks to impose upon Bear Stearns because of its role as underwriter and apparently to read backward onto all of the OS-related claims. Neither does it include “a duty to cause the Trustee to redeem the Bonds on June 1, 1987,” as Eaton Vance contends in its complaint. Bear Stearns undertook to scrutinize certain documents to ensure that certain protections for the bondholders were in place. Since they were not in place, Bear Stearns may be liable under Cal.Civil Code § 3300 for general damages and for those consequential damages which “in the ordinary course of things would be likely to result” from the breach10 —damages incurred because the promised protections were lacking.
This is a relatively limited amount. The surety bond secured $970,000, and the bondholders received $485,000 before the bond expired. This leaves Bear Stearns with $485,000 in total potential liability to all of the bondholders, or about $220,000 in liability to Eaton Vance, which owned roughly 45% of the bonds.11
*14922. Extension of escrow
With respect to the extension of escrow, Eaton Vance has failed to establish a fiduciary duty or any other duty simply as a factual matter. Bear Stearns employees have testified that Eaton Vance authorized them to ask FICAL to extend the deadline. While Eaton Vance argues that a dispute of fact exists on this issue, it has shown only that its representative does not remember having given such permission. This situation does not meet the test of Children’s Television: it is an enormous and unwarranted leap to say that Bear Stearns undertook to act on behalf and for the benefit of Eaton Vance simply by communicating a message to a third party, even if the message was one it was happy to transmit, and even if in that transmission it identified itself as a “representative.” Bear Stearns’ action is far too slight, and not nearly deliberate enough, to bear the consequences with which Eaton Vance wishes to freight it.
Nor is Eaton Vance’s position advanced by the argument that any agreement extending the deadline had to be made by written amendment, and with the unanimous approval of the bondholders. While Bear Stearns did not technically comply with these requirements, it is unclear why that matters. A party does not take on a fiduciary duty just because it violates the terms of a contract.
Finally, Bear Stearns’ letter to the Debt Allocation Committee does not alter the analysis. While the letter does manifest an intention to look after the bondholders’ interests, it is concerned with Bear Stearns’ role in marketing the bonds, not in ensuring that the conditions of the escrow were complied with.
Thus the sweeping liability Eaton Vance seeks to impose is unsupported by the existence of any underlying duty. For this reason, the sweeping remedy it seeks in its seventh claim—rescission—is also without foundation, and we affirm the district court’s summary judgment dismissing that claim. We reverse the dismissal of the sixth claim, but only to the limited extent discussed above.12
Since we affirm the district court’s summary judgment dismissing the first, second, third, fourth, fifth, and eighth claims, this leaves only Eaton Vance’s ninth claim, for mutual mistake. The district court granted summary judgment in favor of Bear Steams on this claim, among other reasons, because it concluded that no mutual mistake had been made. Eaton Vance has not appealed that ruling. Thus we need not reach the parties’ disagreement as to whether the unavailability of rescission as a remedy warrants summary judgment as to the fifth, seventh, and ninth claims, because each of those claims is deficient for separate reasons, or is not be*1493fore us. For the same reason, we do not reach the question of whether the fourth and fifth claims are barred by the statute of limitations.
II
Bondholders v. FICAL13
Background
Eaton Yance, Municipal Securities Trust High Income Series 10, and Municipal Securities Trust High Income Series 11 (collectively, Bondholders) appeal from the district court’s order of summary judgment in favor of FICAL. The Bondholders’ claims arise for the most part from alleged breaches of the Trust Indenture. The major claims are the twelfth, for breaches relating to disbursement of funds, and the fifteenth, for breaches relating to the extension of escrow, the result of which was the avoidance of mandatory redemption.
A. Disbursement (Claim 12)
Under § 4.06(b) of the Trust Indenture, Kress was required to request money from the Construction Fund by means of written requisitions, signed by a Company Representative, and specifying in “reasonable detail” the purpose for which the money was to be used. Further, the trustee was not permitted to make any disbursements until it had “received written notice from the Surety.”
It is not disputed that FICAL failed to comply with at least two of these restrictions on its authority to disburse funds. First, of the 24 requisition certificates submitted by Kress and honored by FICAL, nine were signed by company controller Nancy Zimmerman; the parties agree that she was not a Company Representative as that term is defined in the Loan Agreement. Second, FICAL did not receive the surety’s approval for the disbursements, apparently because FICAL’s trust officers confused the requirement for such approval with the requirement (discussed above) that a surety bond be in place. With respect to the “reasonable detail” requirement, FICAL disbursed more than $4.5 million, often in installments of $500,000, on the strength of requisition certificates which stated simply that the money would be used for “Land and Building.”
The district court granted summary judgment dismissing the twelfth claim. With respect to the signature and detail requirements, the court stated that “there simply is no evidence to show that failure to require more adequate requisitions proximately caused plaintiffs’ losses,” and that “[t]here is no evidence to show that even if ‘reasonable detail’ ... had been given in the requisitions and they had been signed by either Deringer or Klapper that any diversions would have been discovered.” With respect to surety approval, the court concluded that “the surety approval requirement was a typographical error.”
B. Extension of Escrow (Claim 15)
Section 2.06 of the Indenture provided that the bonds were subject to mandatory redemption on June 1, 1987, if the conditions set forth in § 4.06 (requiring Kress to obtain title to parcels of land, a surety bond, and a construction contract) were not met by April 15, 1987. FICAL asserts that representatives from each of the appellant Bondholders consented to extend the deadline by one week. FICAL cites the testimony of Mark Holmstedt, associate director at Bear Stearns, who stated that he had phone conversations with both Peter DeMarco (for Municipal Securities Trust High Income Series 10 & 11) and Daniel Hazard (for Eaton Vance), in which he asked for and was given approval to extend the deadline. Neither DeMarco nor Hazard was able to remember the conversations; both also said that they were not able to recall “one way or the other” whether or not the conversations had taken place.
Bear Stearns then sent a letter to FICAL, stating that Bear Stearns, representing 90% of the bondholders, agreed to extend the escrow period one week. There is no dispute *1494that the deadline was extended, or that FICAL fulfilled whatever obligations it had in assuring that the § 4.06 conditions were fulfilled before the deadline expired.
The district court granted summary judgment dismissing the fifteenth claim on grounds of estoppel and waiver.
C. Other Claims (13, H, and 16)
The Bondholders’ thirteenth claim recasts the alleged breaches of the Indenture as instances of gross negligence. The fourteenth claim recasts them as breaches of fiduciary duty. The district court held that because FICAL had not breached the Indenture, it also had not breached any fiduciary duties nor been grossly negligent.
In their sixteenth claim, the Bondholders allege that FICAL violated the implied covenant of good faith and fair dealing contained in the Indenture by delaying in notifying them that money was being diverted, and by delaying in furnishing them with the documents they requested after they had found out about the diversion themselves. The Bondholders seek only limited relief under the sixteenth claim. After expressing concern that their recovery in other portions of this lawsuit might be reduced on grounds that, by delaying, they failed to mitigate damages, they seek indemnification from FICAL to the extent of any such reduction. The district court held that FICAL had no obligation to produce any documents and that in any event its delay in producing them had not damaged the Bondholders. Discussion
The district court had ancillary jurisdiction over the claims against FICAL. We have jurisdiction under 28 U.S.C. § 1291.
A. Disbursement (Claim 12)
1. Requisition certificates
a. “Reasonable detail”
The district court concluded that the Bondholders had failed to come forward with any evidence which would establish that their losses were proximately caused by FICAL’s failure to require more specific descriptions in the requisition certificates. The court cited to § 8.01(f) of the Indenture, which states that FICAL may “rely upon a certificate signed by a duly authorized representative of the ... Company as sufficient evidence of the facts therein contained” and that “[t]he Trustee may at its discretion secure such further evidence deemed necessary or advisable, but shall in no case be bound to secure the same.” The court reasoned that since FICAL had no obligation to investigate, the Bondholders had failed to show that even if Kress had supplied more detailed requisition requests, the diversions would have been discovered, and the Bondholders would have escaped the damages they suffered.
Although FICAL has adopted this reasoning, it also relies on § 8.01 in a more direct way, arguing that under that provision it did not have any obligation to require any more detailed description than had been supplied.
We reject that argument. Nothing indicates that § 8.01—a general provision—is meant to subsume the more specific requirement for reasonable detail in the requisition certificates. Indeed, principles of construction provide otherwise. See Broad v. Rockwell International Corp., 642 F.2d 929, 947 (5th Cir.) (“A specific provision will not be set aside in favor of a catch-all clause”), cert. denied, 454 U.S. 965, 102 S.Ct. 506, 70 L.Ed.2d 380 (1981). Moreover, § 8.01(f) and the reasonable detail requirement need not be read as contradicting one another: taken together, they indicate that FICAL is entitled to rely on the face of a requisition certificate (rather than being bound to “secure further evidence”) so long as the face of the document contains “reasonable detail.” FI-CAL is not entitled to summary judgment on the basis that § 8.01 negates the reasonable detail requirement.14
*1495Turning to the more serious issue of causation, it is useful first to discuss the terminology used in this area, as it is often employed in a confusing way under California law. Both the parties and the district court labelled the causation issue in this case “proximate cause.” Their reasoning, however, focuses on cause in fact. FICAL argues that “the same loss would have occurred” whether or not Deringer had submitted inadequate requisitions. Br. of FICAL at 30. Similarly, the district court reasoned that “even if’ adequate requisition certificates had been supplied, Deringer’s misdeeds would not have been discovered, and plaintiffs’ injuries would have occurred. This language reveals that the causation question which the parties have litigated concerns whether or not the alleged breaches were the “but for” cause, or cause in fact of injury.
The parties’ and the district court’s confusing use of the term proximate cause to describe the concept of cause in fact is understandable, given the state law context in which they were operating. In Mitchell v. Gonzales, 54 Cal.3d 1041, 1 Cal.Rptr.2d 913, 819 P.2d 872 (1991), the California Supreme Court pointed out that the term “proximate cause” is inappropriately used in California jury instructions to refer to cause in fact, id. at 914 n. 2, 819 P.2d at 873 n. 2; see also Helm v. K.O.G. Alarm, Co., Inc., 4 Cal.App. 4th 194, 5 Cal.Rptr.2d 615, 619 n. 8 (1992), and Mitchell’s discussion of the caselaw in this area reveals that the California courts have at times perpetuated the confusion. Nevertheless, Mitchell emphasized that true proximate cause—what the California jury instructions call “legal cause”—is something different from, and something more than, cause in fact. 1 Cal.Rptr.2d at 914 n. 2, 819 P.2d at 873 n. 2; Hardison v. Bushnell, 18 Cal.App.4th 22, 22 Cal.Rptr.2d 106, 107 (1993).
The California courts have recognized that there are two ways in which a plaintiff can show cause in fact: by showing that defendant’s conduct was the “but for” cause of injury, or by showing that it was a “substantial factor” in that injury. Mitchell, 1 Cal.Rptr.2d at 914 n. 2, 919-20, 819 P.2d at 873 n. 2, 877-78; Brookhouser v. State, 10 Cal.App.4th 1665, 13 Cal.Rptr.2d 658, 665 (1992) (“There are two widely recognized tests for determining whether a defendant’s conduct has in fact caused the plaintiffs injury: Whether the injury would not have occurred but for the defendant’s conduct, and whether the defendant’s conduct was a substantial factor in bringing about the injury.... Mitchell ... does not appear to alter these two tests for cause in fact”) (emphasis in original).15 In Mitchell, the Supreme Court also recognized that the “but for” requirement is subsumed in the substantial factor test, since an action cannot be a substantial factor in an injury (or any kind of factor at all) if it had nothing to do with bringing that injury about. 1 Cal.Rptr.2d at 920, 819 P.2d at 878-79.16 The latter principle was recognized by the California Supreme Court long ago. In Signorelli v. Potter, 43 Cal.2d 541, 545, 275 P.2d 449 (1954), the court held that if a plaintiffs injury would have come about even in the absence of defendant’s breach, then that breach was not a substantial factor. Signorelli, 43 Cal.2d at 545, 275 P.2d 449.
*1496It is on the basis of this latter principle that FICAL argues the district court’s ruling should be upheld. FICAL’s argument is that even if Deringer had submitted more detailed requisition certificates, the bondholders’ injury would not have been averted because FICAL, which was entitled to rely on the face of the documents, would not have discovered that Deringer was lying about the purpose to which the funds were being put. The question here, like the question in Signo-relli, is whether, in the absence of the breach, injury would still have occurred.
In answering the question positively, FICAL omits a crucial step in the causal chain. The correct question is not what would have happened if Deringer had submitted properly detailed certificates, but rather what would have happened if FICAL had not breached its duty17 — if, instead of disbursing funds in contravention of the Indenture, it had refused to do so. As an initial matter, the answer is that the Bondholders would have escaped injury if this had been the case. But for the breach — disbursement on the basis of inadequate requisition certificates — Deringer would not have been able to squander the Bondholders’ funds because he would not have gotten those funds in the first place.
Following through with this necessarily speculative exercise, however, we must next ask what Deringer might have done had FICAL refused to let him have the money. He might have supplied the missing detail and gone on to divert funds in exactly the same manner he actually employed.
This was the district court’s conclusion. But having added a missing link to the causal chain, it becomes clear that this is not the only alternative course of events which might have occurred in the absence of a breach. Deringer might also have been deterred; he might have concluded that with a vigilant trustee, his wrongdoing would be detected. He might have balked at supplying “reasonably detailed” lies about the destination of the requisitioned funds: although FICAL had no duty to investigate, it clearly had the right to do so under § 8.01, and Deringer might have feared, had the bank indicated to him that it was on notice of the irregularities in his requests, that it would decide to look into the matter further. If FICAL had held Deringer to the terms of the Trust Indenture, there is some chance that he would have decided that the Long Beach project was not a good candidate for diversion.
This is necessarily speculative. As Justice Traynor recognized in Signorelli “[ojrdinarily it cannot be proved conclusively what would have happened if something else had not happened.” 43 Cal.2d at 546, 275 P.2d 449. But it is also speculative to conclude that Deringer would have simply provided the missing details and diverted the money anyway. Choosing between the speculations is ordinarily a question for the trier of fact, who must “determine the balance of probabilities.” Signorelli at 546, 275 P.2d 449. Thus far, summary judgment is inappropriate.
In the district court, however, FICAL suggested that the deterrent theory relied on by the Bondholders adds yet another layer of speculativeness to what is already a speculative inquiry, and by doing so renders plaintiffs’ case so speculative that it should not go to trial. Analogizing the reasonable detail requirement to a different kind of deterrent device — a burglar alarm — FICAL cited to Guthrie v. Am. Protection Indus., 160 Cal.App.3d 951, 206 Cal.Rptr. 834 (1984), in which the court had to decide whether or not to honor a liquidated damages clause in a contract between the buyer and the seller of a burglar alarm system. In upholding the clause on the basis of “extreme difficulty in fixing actual damages,” the court stated that “[t]he success of [an alarm] system depends on many variables and intangibles.... no one really knows what is ‘adequate’ deterrence in any given situation.... it is our opinion that it would be impossible in any case to prove, after the fact, that an operative alarm system would have prevented the *1497crime.” Id. 206 Cal.Rptr. at 836 (internal citations omitted).
Guthrie, however, has subsequently been disapproved. See Helm, 5 Cal.Rptr.2d at 618-19. More importantly, in the case before us, and in contrast to both Guthrie and Helm, it is actually quite simple to determine, as an initial matter, whether or not the Bondholders would have suffered losses in the absence of a breach. If FICAL had not disbursed funds on the basis of the inadequate requisition certificates that Deringer was submitting, Deringer would not have gotten the money and could not have diverted it. In order to show any different causal relationship between breach and damages, it is FICAL which must begin to speculate, and which must imagine what might have happened in the absence of breach. This is different from the burglar alarm eases, where it is the plaintiff who must begin speculating by asking what would have happened if the device had gone off. Here, the Bondholders’ speculations about deterrence are made only to refute FICAL’s initial speculations — which are made to combat the Bondholders’ successful demonstration of cause in fact as a matter of simple logic. The difficulty in conclusively proving deterrence must be viewed differently in a case where speculation is initiated by the defendant’s attempt to refute an initial showing of cause in fact, rather than by the plaintiffs attempt to show that element in the first place. The difficulty becomes the defendant’s problem.
This logical difference between Guthrie and the present ease corresponds to a difference in the real world. While the burglar alarm in Guthrie was simply a deterrent, here the breaching party played two roles— both an affirmative and enabling role (giving Deringer the money when all conditions in the Indenture were met) and a deterrent role (by enforcing the conditions to the extent that it was required to do so). FICAL’s breach, in the first instance, was in affirmatively continuing to enable Deringer to squander the funds. Only secondarily does the breach consist of a failure to deter — and does the Bondholders’ causation case become speculative. Stated differently, when an entity whose business it is to facilitate the actions of a person who turns out to be a wrongdoer breaches a duty to others in its dealings with that person and by doing so facilitates his wrongdoing, that entity may have caused the ensuing injury in a way in which the entity whose breach merely allows a wrong to occur has not. A bank is not a burglar alarm; it has positive as well as negative functions, and as a result its breaches of duty can cause harm in more ways than one.
We now return to the district court’s order, in which the court stated that there was “no evidence to show that failure to provide more adequate requisitions proximately caused plaintiffs’ losses,” and that the Bondholders had made “no showing that Kress could not have satisfied all of the requisition requirements ... and still not [sic] have diverted the bond funds” (emphasis added). The Bondholders’ theory of causation — the deterrence theory just outlined — is not evidence. But by the same token, FICAL also has provided no evidence on the question of causation — just its own assumptions about what Deringer would have done if FICAL had not breached.18
We must determine whether the district court was correct in granting summary judgment when there appeared to be no evidence from either side. The case which addresses summary judgment motions made in this posture is Celotex Corp. v. Catrett, 477 U.S. 317, 106 S.Ct. 2548, 91 L.Ed.2d 265 (1986). There, the Supreme Court squarely rejected the idea that the party moving for summary judgment must itself produce evidence if it is to show the absence of a genuine issue of *1498material fact. Rather, the Court held, the moving party’s burden on summary judgment “may be discharged by ‘showing’ — that is, pointing out to the district court — that there is an absence of evidence to support the nonmoving party’s case.” 477 U.S. at 325, 106 S.Ct. at 2554. Under this analysis, the Bondholders’ failure to produce any evidence initially seems fatal.
Celotex’s allocation of summary judgment burdens, however, applies only when it is the nonmoving party which would have the burden of proof at trial. 477 U.S. at 325, 106 S.Ct. at 2554. In this case, while the Bondholders do have the burden of proving cause in fact,19 they have already shifted that burden: they have shown cause in fact simply by pointing to the fact that Deringer’s misdeeds would have been impossible had FICAL not breached the Indenture by disbursing funds. Evidence about deterrence becomes important only in light of FICAL’s attempt to rebut that initial showing. In this situation, the parties’ joint failure to come forward with evidence about deterrence at the summary judgment stage is properly taxed to defendant rather than to plaintiff.20 We therefore conclude that the district court erred when it granted summary judgment to FICAL on the grounds that the Bondholders had failed to produce evidence showing that their injuries were caused by FICAL’s non-enforcement of the reasonable detail requirement.
b. Company Representative signature
The Bondholders also argue that FI-CAL’s disbursements violated the terms of the Indenture because nine of the requisition certificates were not signed by a Company Representative, as required by § 4.06. Once again the issue is whether that breach caused the Bondholders’ injuries. Our analysis here is simpler than it was with respect to the reasonable detail requirement, because both parties submitted evidence on the issue of causation.
FICAL has shown that money drawn from the Construction Fund by means of the nine requisition certificates signed by Zimmerman was put into the same Deringer-controlled bank account as was the money drawn by means of certificates that Deringer himself had signed. More fundamentally, there is no dispute that Deringer did sign the other 15 certificates. The fact that he was willing to do so, FICAL argues, shows that he would not have been deterred if FICAL had refused to honor the certificates signed by Zimmerman and had insisted on his signature instead.21
Nevertheless, the Bondholders reiterate their deterrence argument in this context, and suggest that if FICAL had enforced the provision, Deringer might have decided that it was too risky to carry out his scheme. In support of this theory, the Bondholders contend that when FICAL finally did discover the signature problem and the bank’s trust officer asked Zimmerman to have Deringer validate her signature on the requisitions, apparently retroactively, Deringer never did *1499so. The argument therefore is that Deringer would have been deterred if FICAL had enforced the requirement.
The value of this evidence could easily be overstated. In declining to authorize the signatures retroactively, Deringer refused to take an action which would have put him at risk when there was no longer anything to gain, since he already had the money. This does not necessarily prove that he was subject to deterrence. Nevertheless, the evidence is sufficient to create a dispute of fact on the question of causation. Thus summary judgment was inappropriate with respect to the Company Representative signature requirement.
2. Surety approval
There is no dispute that FICAL gave money to Kress Associates without once receiving “written Notice from the Surety,” notwithstanding the fact that § 4.06 of the Indenture listed this as a prerequisite. FI-CAL contends that the district court was correct in concluding that the surety approval clause was a typographical error. FICAL cites to the deposition testimony of Andrew C. Hall, bond counsel for the City and drafter of the Indenture. Hall testified that retaining the clause was “an error,” that “the whole phrase should have been stricken,” and that “in my mind I don’t think there’s any question but that that was an error.” Similarly, Stephen Pochos, counsel for Kress, testified that the phrase “should have been taken out.” Pochos explained that it was a remnant from a different indenture agreement which had been used as a template for the Kress Project Indenture (and in which the surety had played a larger role), and that at a meeting among various parties to the bond transaction, a FICAL representative identified dozens of such references, the rest of which apparently were removed.
To refute this evidence, the Bondholders state that Hall, the drafter, also testified that he had no specific recollection as to whether the surety approval requirement was a typographical error. The Bondholders argue that this creates a factual dispute on the question of whether or not a mistake had been made in drafting—all the more so because under California law, evidence of mistake must be “clear and convincing” if it is to justify reformation. Matter of Beverly Hills Bancorp, 649 F.2d 1329, 1333 (9th Cir.1981) (summarizing California law).
This argument is unpersuasive. In the deposition testimony the Bondholders cite, Hall stated only that he could not remember whether in December 1986 he understood the surety approval clause to be a typographical error. This does not conflict with the testimony FICAL cites, in which Hall said conclusively that the phrase was retained erroneously. Since Hall’s two statements are not inconsistent with one another, there is no dispute of fact on this issue. Neither does the “clear and convincing” standard help the Bondholders: both Hall and Pochos stated quite positively that the retention of the surety approval clause was an error, and Pochos provided an explanation for its presence in the document.22
Pochos’ explanation also lays to rest another challenge raised by the Bondholders. The Bondholders complain that FICAL has not explained why it did not catch the error before the Indenture was executed; in support of the contention that such explanation is required, they cite Appalachian Ins. v. McDonnell Douglas Corp., 214 Cal.App.3d 1, 262 Cal.Rptr. 716 (1989). The court in that case, however, sought such an explanation not because it required the party seeking reformation to exculpate itself, but rather as evidence that what the party claimed to be a mistake was in fact a mistake. The court sought to have the party seeking reformation show that it had been negligent in its preparation of the document rather than that it had not, since evidence of negligence would *1500have shown that the mistake was made at the time of drafting, and would have alleviated the court’s concern that the party was trying to get out of a contract it no longer liked by claiming, ex post, that a clause which had been written quite deliberately was actually an error. 262 Cal.Rptr. at 727. The explanation the Appalachian Insurance court sought has been introduced in this case: Poc-hos explained the clause as an inadvertently retained leftover from a previous indenture agreement, in which the surety had played a greater role.
Appalachian Insurance also states in dictum that reformation may be denied because a party has been negligent. Id. 262 Cal. Rptr. at 726. But the operative word is “may”: a court may decline to grant the reformation request of a negligent party for equitable reasons. Id. Here, we do not see how equity would be disserved by allowing reformation of a phrase which every person who testified on the issue agreed was a mistake. The Bondholders have not shown that they relied on the surety approval clause, and thus they are not being denied the benefit of anything they bargained for.23
The Bondholders’ remaining challenges to reformation are similarly unpersuasive. They suggest that reformation is not appropriate because the FICAL officer in charge of their account did not believe that the clause was a mistake; rather, he misunderstood it entirely, and confused it with the requirement for a surety bond. However, the Pochos deposition, referred to above, indicates that the FICAL representative involved in drafting the document did intend conditions like the surety approval clause to be removed. Finally, the Bondholders argue that reformation is procedurally improper, because it was not listed as an affirmative defense. However, the Bondholders were put on notice that a typographical error defense would be raised, at the latest, when FICAL submitted the Statement of Uncon-troverted Facts which accompanied its summary judgment motion. The Bondholders presented a vigorous response to that defense in the district court. Under such circumstances, it would be hyperteehnical to deny FICAL the benefit of the defense. See, e.g., Albert v. Joralemon, 271 F.2d 236, 241 (9th Cir.1959).
B. Extension of Escrow (Claim 15)
1. Dispute of fact as to assent
With respect to the fifteenth claim, the Bondholders argue that summary judgment—based on waiver and estoppel—was inappropriate because there was a genuine dispute of fact as to whether or not they assented to extending the escrow period. The Bondholders argue that Hazard’s and DeMarco’s testimony that they could not recall having conversations in which they gave their consent conflicts with Holmstedt’s testimony that such conversations did occur. The Bondholders cite Wigmore for the proposition that evidence based on “negative knowledge” is admissible:
there is no inherent weakness in this kind of knowledge. It rests on the same data of the senses [as testimony that an event did occur]. It may even sometimes be stronger than affirmative impressions. The only requirement is that the witness should have been so situated that in the ordinary course of events he would have heard or seen the fact had it occurred. This sort of testimony is constantly received—particularly in proof of ... negative facts.
2 Wigmore on Evidence § 664 (Chadbourn rev. 1979) (emphasis in original); see also Morris v. Towers, 199 Cal.App.2d 701, 19 Cal.Rptr. 135, 137 (1962).
The “negative knowledge” principle, however, is of no assistance to the Bondholders. Hazard and DeMarco did not testify to the absence of some sensory impression; rather, they stated that they could not remember one way or the other whether a conversation had occurred. The inference to be drawn *1501from the absence of a sensory impression is that the event which was not seen-or heard did not occur. But the same inference cannot be drawn from the testimony that a person does not remember what happened. The situation would have been different if Hazard and DeMarco had said that they had not heard Holmstedt asking for their consent, or more to the point, if they had said that they did not give such consent. This could have supplied crucial proof of a “negative fact” — the absence of assent. But their testimony was that they do not remember what happened, and that does not prove anything.24
The Bondholders also argue that a dispute of fact exists on the issue of consent because Holmstedt’s credibility is questionable. They cite Lodge Hall Music, Inc. v. Waco Wrangler Club, Inc., 831 F.2d 77, 81 (5th Cir.1987), in which the court implicitly held that evidence which calls into question a witness’s credibility with respect to one statement is sufficient to raise a factual issue about the credibility of other, related statements made by the same witness. The Bondholders point to uncertainty surrounding the letter communicating the assent to extend escrow, which bore what purported to be Holmstedt’s signature. It is unclear who drafted the letter, who signed it, when it was written, and when it was received. But none of this impeaches Holmstedt’s credibility. Holmstedt himself testified that he did not draft or sign the letter, and he did not say that it was written on any particular date. Although uncertainty surrounds the letter, the credibility of Holmstedt’s testimony about the letter has not been called into question. Thus Holmstedt’s testimony about the letter does not taint his testimony about obtaining assent. In short, the Bondholders have not raised a dispute of fact as to whether or not assent was given.
2. Elements of estoppel and waiver
The Bondholders also argue that summary judgment dismissing their fifteenth claim was inappropriate because FICAL failed to establish several of the elements of a waiver or estoppel defense. Most basically, the Bondholders argue that neither defense is available to FICAL because FICAL has not shown that the Bondholders’ assent was ever communicated to the Bank. The Bondholders’ factual basis for this argument is a footnote in FICAL’s response brief, in which FICAL concedes that “there may be factual issues as to the exact date of th[e] letter” from Holmstedt to Frank Sulzberger at FICAL, and then goes on to state that these issues are unimportant because “FICAL’s Motion for Summary Judgment was not based on that letter.” Br. of FICAL at 36 n. 23. The Bondholders seize upon the statement that the motion was not “based” on the letter to conclude that FÍCAL has admitted that the letter was never sent.
That conclusion lacks support. FICAL contended below that the Bondholders’ assent was communicated to the bank, and the Bondholders never argued that it was not.25 We will not permit the Bondholders to create a dispute of fact by misreading a footnote in an appellate brief.
Next, the Bondholders argue that even if the assent was communicated, there is a dispute of fact as to whether the communication was timely, since it has not been established that it was received before April 15. They argue that waiver must occur “before expiration of the time originally fixed for performance.” 5 Williston on Contracts § 689 at 306 (3d ed. 1961). However, the Bondholders are confusing the time for FICAL’s performance with the time for Kress’s performance: it was Kress which had to *1502meet the § 4.06 conditions by April 15, 1987. The time for FICAL’s performance was June 1, 1987, when the bonds were to be redeemed. The Bondholders do not argue that waiver was not made before this later date.
Next, the Bondholders cite Perini v. Perini, 225 Cal.App.2d 899, 37 Cal.Rptr. 354, 359 (1964), for the proposition that waiver requires “full knowledge of the facts.” The facts which were required to be known in Perini, were facts necessary for an understanding of the rights which were waived— there, facts pertaining to divorce law. See also In re Marriage of Moore, 113 Cal.App.3d 22, 169 Cal.Rptr. 619, 621 (1980) (“There must be actual or constructive knowledge of the existence of the right to which the person is entitled”). Here, there is no apparent connection between the facts the Bondholders suggest were omitted—facts about why the extension was needed and which specific conditions of the escrow had not been met—and the right the Bondholders were waiving. Thus this argument fails too.
Finally, the Bondholders contend that estoppel requires reasonable reliance and that FICAL’s reliance on the representation that the Bondholders had consented to an extension of the deadline was not reasonable in light of the fact that amendments to the Indenture had to be approved by 100% of the bondholders and had to be in writing. But this does not show that FICAL’s reliance was unreasonable: it is not unreasonable to conclude that someone who is waiving a substantive right (the right to mandatory redemption) is also waiving a procedure (a specific method for amending an agreement). The Bondholders’ final argument fails. We affirm the district court’s grant of summary judgment dismissing the Bondholders’ fifteenth claim.
C. Other Claims (IS, U, and 16)
In dismissing the Bondholders’ thirteenth claim (for gross negligence) and fourteenth claim (for breach of fiduciary duty), the district court cited to § 8.01(a) of the Indenture, which states that “[t]he Trustee, prior to the occurrence of an Event of Default and after the curing of all Events of Default which may have occurred, undertakes to perform such duties and only such duties as are specifically set forth in this Indenture” (emphasis added). The court concluded that in light of this clause, the Bondholders could show gross negligence and breach of fiduciary duty only if they also showed breaches of the Indenture.
As to the thirteenth claim, FICAL does not argue that the Bondholders’ evidence failed to meet the standard for gross negligence, but only that the Bondholders failed to show that any breach of the Indenture proximately caused their injuries. To the extent that we rejected that contention in the context of the twelfth claim, we also reject it in the context of the thirteenth, and we reverse in part the district court’s grant of summary judgment as to that claim. We also note that the Indenture itself provides that FICAL may be “answerable” for gross negligence in the performance of its duties. Indenture at § 8.01(g).
We affirm summary judgment with respect to the fourteenth claim. When a limiting clause such as § 8.01(a) is included in an indenture of trust, the trustee’s obligations are confined to the scope of the contract, and do not extend to common law fiduciary duties. See supra at 10701-10703 & n. 9.
FICAL argues that summary judgment dismissing the Bondholders’ sixteenth claim should also be upheld on the basis of § 8.01(a) of the Indenture: FICAL contends that since no duty to provide documents to bondholders is contained in the Indenture, FICAL cannot be hable for any alleged breach of such a duty. The Bondholders rely on § 8.01(a) of the Indenture, which states that “[i]n case an Event of Default has occurred (which has not been cured or waived), the Trustee shah ... use the same degree of care and skill in the exercise [of its rights and powers under the Indenture] as a prudent man [sic] would exercise or use under the circumstances in the conduct of his own affairs.” The Bondholders argue that an Event of Default occurred when the conditions of § 4.06 were not met by April 15, 1987.
*1503We are not persuaded that Kress’s failure to timely meet the § 4.06 conditions amounted to an “Event of Default.” Section 7.01 of the Indenture, which defines that term, states that it includes failure to make payments which are due, default by the City in performance of its obligations, and continuing default under the Loan Agreement. The problems Kress experienced at closing do not fall under any of these categories. But even if there had been an Event of Default, this would not trigger the post-default duties the Bondholders seek to impose. The clause the Bondholders rely on specifies that those duties apply only if the default “has not been cured or waived.” By assenting to an extension of the escrow, as discussed above, the Bondholders waived the purported default.
In the alternative, the Bondholders look outside the Indenture for the source of the higher level of duty they seek to impose under the sixteenth claim. They rely upon the implied covenant of good faith and fair dealing, which, under California law, requires one party to refrain from doing anything which “injures the right of the other to receive the benefits of the agreement,” Bleecher v. Conte, 29 Cal.3d 345, 213 Cal.Rptr. 852, 854, 698 P.2d 1154, 1156 (1981) (internal quotation marks omitted), and more, positively, requires the parties to carry out existing contractual obligations fairly and in good faith. Koehrer v. Superior Court, 181 Cal.App.3d 1155, 226 Cal.Rptr. 820, 828 (1986). But the eases cited by the Bondholders do not hold that the covenant requires a party to take on additional duties—such as the duty to produce documents, or to inform investors of possible disasters which have befallen the project in which they have invested. Indeed, several courts have stated explicitly that the covenant of good faith and fair dealing does not alter specific obligations set forth in the contract, Balfour, Guthrie & Co., Ltd. v. Gourmet Farms, 108 Cal.App.3d 181, 166 Cal.Rptr. 422, 427 (1980), or add duties independent of the contractual relationship. Moore v. Local 569, Intern. Broth, of Elec. Workers, 653 F.Supp. 767, 774 (S.D.Cal.1987) (applying California law). We therefore affirm summary judgment dismissing the Bondholders’ sixteenth claim.
Ill
FICAL v. Brobeck, etc., et al.
Background
The law firm Brobeck, Phleger & Harrison served as counsel for Bear Stearns, while the law firm Jones, Hall, Hill & White was counsel for the City. Both firms were defendants in the Shawmut action, in which the Bondholders sued them for fraud, negligent misrepresentation, negligence, and violation of § 10(b). All of the Bondholders’ claims against the law firms arose from alleged misrepresentations in the offering documents, and were part of the securities fraud portion of this litigation, discussed in the first part of this opinion. In the trust-administration portion of the litigation, discussed in the second part of this opinion, FICAL cross-claimed against the law firms, alleging that the firms were hable for indemnity and contribution with respect to claims brought against FICAL.
Like Bear Stearns, the law firms filed a summary judgment motion in the district court based in part on plaintiffs’ purported inability to show loss causation. The district court concluded that plaintiffs’ losses were not caused by the law firms or by Bear Stearns. The law firms moved for entry of judgment based on district court’s conclusions of law, but before that motion was heard, the firms decided to settle with plaintiffs, each firm paying $400,000. The firms then moved for a determination that their settlement had been made in good faith, and for an order from the court dismissing ah claims for contribution and indemnity filed against them by any party in the lawsuit. FICAL opposed the motion, arguing that the settlement did not take account of its cross-claims against the law firms.
The district court ruled on the law firms’ good faith settlement motion in the same memorandum order in which it granted summary judgment for Bear Stearns on the three rescission claims and summary judgment for FICAL on the five trust-administration claims. The court stated that
The only party that opposes Brobeck and JHHW’s motion is FIB [FICAL], Be*1504cause the court grants summary judgment in favor of FIB in Motion # 3, this motion is granted.
In the conclusion section of the order, the court stated that “Brobeck and JHHW’s good faith settlement motion is granted.”
FICAL cross-appeals from this order, pointing out that if we reverse the summary judgment order entered in its favor—which we do in part—it will be exposed to liability without the benefit of possible contribution and indemnification from the law firms. The law firms argue that the district court’s order approving the settlement should be affirmed. Discussion
A. Interpretation of the District Court’s Order ■
FICAL argues that the district court erred because it incorrectly assumed that by granting FICAL’s summary judgment motion, it rendered FICAL’s opposition to the law firms’ settlement motion moot. The parties do not dispute that if the district court did in fact rely on this assumption, its ruling was erroneous. The parties’ disagreement is over the proper interpretation of the district court’s one-sentence disposition of the law firms’ motion, “Because the court grants summary judgment in' favor of FI[CAL], this motion is granted.” FICAL reads the sentence to mean “Because FICAL has prevailed on summary judgment, its opposition is moot, and the law firms’ motion is granted as unopposed.” The law firms read the sentence to mean “Because the claims which were the subject of the summary judgment motion were fatally weak, rendering summary judgment appropriate, it was reasonable for a settlement agreement not to take account of possible liability for indemnification based on those claims; and because the settlement was therefore reasonable, the law firms’ motion is granted on the merits.”
The law firms’ reading of the district court’s order seems strained, and certainly FICAL’s is the more parsimonious of the two. But we need not choose between them. If a trial court’s findings of fact are such that it is “impossible ... to determine whether the district court properly applied the law to the facts,” the appellate court may remand for more specific findings. Norris v. City and County of San Francisco, 900 F.2d 1326, 1329 (9th Cir.1990).26 Given the ambiguity in the district court’s order, we choose to remand. The district court may re-enter its order approving the settlement if it did in fact originally make a good faith settlement determination on the merits; otherwise, the district court must vacate the order. If it vacates the order, the law firms are free to re-file their motion and allow the district court to decide that motion on the merits— taking into account, of course, the way in which the landscape has changed as a result of our partial reversal of the summary judgment in favor of FICAL.
B. Alternative Bases for Affirmance
We reject the law firms’ invitation to affirm on the merits regardless of what the district court’s basis for its ruling may have been, although we acknowledge that we may affirm the district court on any basis supported by the record. United States v. Washington, 969 F.2d 752, 755 (9th Cir.1992), cert. denied, — U.S.-, 113 S.Ct. 1945, 123 L.Ed.2d 651 (1993). A good faith determination under Cal.Code Civ.Proe. § 877.6 is both fact-sensitive and discretionary, Tech-Bilt, Inc. v. Woodward-Clyde & Assoc., 38 Cal.3d 488, 213 Cal.Rptr. 256, 263-65, 698 P.2d 159, 166-68 (1985), and we must avoid finding facts in the first instance. See, e.g., Norris, 900 F.2d at 1332 (appellate court remands where district court’s findings of fact are unclear). The district court’s order may be upheld on alternative grounds only if the good faith determination was mandated as a matter of law. The law firms have not shown that it was.
*15051. No duty
FICAL’s cross-claims are based primarily on the law firms’ role in drafting the trust documents: FICAL argues that the law firms breached duties owed to FICAL by negligently drafting the Indenture, such that, among other things, it required reasonable detail in the requisition certificates without defining what that term meant. The law firms argue that the good faith determination should be affirmed because as a matter of law they had no duty to FICAL, which was not their client.
The eases on which the law firms rely, however, do not bear this out. The law firms argue that St. Paul Title Co. v. Meier, 181 Cal.App.3d 948, 226 Cal.Rptr. 538 (1986), is directly on point. In that case, the purchaser in a real estate transaction sued an escrow agent for disbursing funds contrary to the purchaser’s instructions. The agent cross-claimed against the purchaser’s attorney, who had drafted the instructions for the purchaser. The appellate court affirmed dismissal of the cross-claim.
In Meier, the cross-claimant “[n]ever met, consulted with, received or relied upon any legal advice or opinion from respondent attorney.” 226 Cal.Rptr. at 539. Since the attorney and the third party were therefore not in privity with one another, the court concluded that liability could be imposed only if the third party was “the intended beneficiary of the attorney’s services, or the foreseeability of harm to the third party as a consequence of professional negligence [was] not outweighed by other policy considerations.” Id. Neither requirement was met.
In this case, by contrast, there is some question as to whether or not FICAL is properly considered a third party at all, or in any event whether it was an “intended beneficiary.” FICAL contends that the law firms undertook drafting and proofreading responsibilities on behalf of all parties to the bond transaction.27 That contention may be questionable in light of the fact, emphasized by the law firms, that FICAL was represented by its own counsel throughout the bond transaction. But this raises factual issues, and thus takes us outside of the realm in which we may find alternate grounds for affirmance.
The law firms also rely on Bily v. Arthur Young and Co., 3 Cal.4th 370, 11 Cal.Rptr.2d 51, 834 P.2d 745 (1992), primarily for its reference to Goodman v. Kennedy, 18 Cal.3d 335, 134 Cal.Rptr. 375, 556 P.2d 737 (1976). The latter case made it clear that lawyers ordinarily owe no duty to third persons with whom their clients deal at arm’s length. Id. at 381, 556 P.2d at 743. But again, whether or not “arm’s length” properly captures the relationship among the various attorneys and principals in this case is a question of fact. We decline to affirm on the ground that, as a matter of law, the law firms had no duty to FICAL.
2. No breach of duty
The law firms also argue that they are entitled to a favorable good faith determination as a matter of law because they gave no advice about the parts of the transaction on which FICAL’s alleged liability is based. But once it is accepted that the firms may have had a duty to exercise reasonable care even though FICAL was not their client, this argument is unpersuasive, because the firms clearly were involved in those parts of the transaction. Thus at least one possible breach of the law firms’ duties is the failure, on the part of both, to define reasonable detail in a way which would make the requirement clear and easy to apply.
We express no opinion about the relative strength of FICAL’s claims that the law firms breached their duty in this or any of the other ways set forth in FICAL’s brief. Determining FICAL’s chances of success on those claims, together with balancing numerous other litigation-specific factors and broader policy considerations, is the job of the district court in the first instance. We remand so that the district court may perform that function.
*1506Each party shall bear its own costs on appeal.
AFFIRMED IN PART, REVERSED IN PART, and REMANDED with respect to Eaton Vance’s claims against Bear Stearns.
AFFIRMED IN PART, REVERSED IN PART, and REMANDED with respect to the Bondholders’ claims against FICAL.
REMANDED with respect to FICAL’s cross-claims against Brobeck, Phleger & Harrison and Jones, Hall, Hill & White.
. Because Eaton Vance admitted that it never read the OS, a fair question is whether we should look to that document in assessing whether actionable misrepresentations were made. Ultimately the question is whether Eaton Vance reasonably relied on any offering document. We need not resolve that question, however, since we conclude that the OS contained no actionable misrepresentations. We refer to the OS rather than the POS as the source of the alleged misrepresentations because we conclude that at the very least, Eaton Vance did not reasonably rely on the POS, which was stamped "draft" on nearly every page; therefore, at a minimum, knowledge of changes made in the OS must be imputed to Eaton Vance. We reiterate, however, that we do not decide that Eaton Vance reasonably relied on any document or composite of documents.
. The "track record” claim fails for another reason as well. Eaton Vance does not contend that Bear Steams knew about the intercompany transfers, but only that, through the exercise of due diligence, it should have known about them. But a plaintiff suing under § 10(b) must show scienter, i.e., knowing or reckless misconduct. Ernst & Ernst v. Hochfelder, 425 U.S. 185, 96 S.Ct. 1375, 47 L.Ed.2d 668 (1976); Hollinger v. Titan Capital Corp., 914 F.2d 1564 (9th Cir.1990), cert. denied, 499 U.S. 976, 111 S.Ct 1621, 113 L.Ed.2d 719 (1991); McGonigle, 968 F.2d at 817; Apple, 886 F.2d at 1116-17. While a due diligence requirement may be read into § 12(2) of the 1933 Act, in which the standard is negligence, that requirement may not be relied on by municipal bondholders confined to 1934 Act remedies. 15 U.S.C. § 771(2); see In re New York City Mun. Sec. Litig., 507 F.Supp. 169, 178-79 & n. 20 (S.D.N.Y.1980).
. Eaton Vance's argument on this point is somewhat shifting. Eaton Vance also suggests that the OS was misleading because it did not reveal that there was a need for funds in excess of $9,965,000; Eaton Vance suggests that this fact was concealed because Bear Steams believed that the federal tax exemption was available only if the total expenditures appeared to be less than $10,000,000. This was the argument made to the district court. It was rejected there because the court concluded that a careful reading of the OS did reveal that over $11,000,000 was needed.
Alternate grounds for rejecting this argument exist as well. Unlike the OS, the POS quite explicitly stated that there was a $2,000,000 difference between the bond proceeds and the amount needed to complete the project. Although we explained above that we look to the OS rather than the POS, see supra note 1, reference to the POS is nevertheless warranted in this instance, where the POS is accurate and only the OS is arguably in error. Because Eaton Vance should have read the OS, knowledge of the changes made therein may be imputed to it. But it does not follow that the facts in the POS—the document Eaton Vance actually read—can be ignored. Thus Eaton Vance may not argue, on the basis of the OS, that it did not know of the shortfall.
.Indeed, correspondence enclosed with the letter from Madden to Dashiell indicates just the opposite.
. Holmstedt also testified that both he and the attorneys involved in the bond issue understood the guaranty in a more limited way. The lawyer who wrote the document, however, could not recall whether or not, at the time she drafted it, she understood it as terminating only at the time the debt service coverage test was met, or whether it would expire upon default. She also could not recall whether she thought the OS indicated that the guaranty would expire in the event of default and acceleration.
. We also note that as the transaction subsequently unravelled, the terms of the guaranty became essentially irrelevant. Kress Associates is a partnership. Once its assets were exhausted, bondholders and other creditors could have sought satisfaction from the assets of the individual partners, Deringer and Klapper, whether or not a guaranty entitled them to do so.
. Eaton Vance does continue to argue that Bear Stearns has a fiduciary duty under the Investment Company Act of 1940, 15 U.S.C. §§ 80a-l, et seq., because two of the bondholders, the Municipal Securities Trust High Income Series 10 & 11, are its own unit trusts. The unit trusts, however, have not sued Bear Stearns (although they have sued FICAL and others in the related action), and Eaton Vance does not explain how it has standing to litigate a purported breach of a duty owed to another. The fact that, under the Investment Company Act, Bear Steams may have had a duty to treat its own institutional purchasers no better than any other purchasers does not change the analysis, because Eaton Vance does not assert that Bear Stearns breached this duty by treating it any differently than it treated the unit trusts.
Eaton Vance also cites, though without specifically promoting, California cases which have held that a stockbroker owes a fiduciary duty to a customer. Blankenheim v. E.F. Hutton & Co., 217 Cal.App.3d 1463, 266 Cal.Rptr. 593, 601 (1990); Duffy v. Cavalier, 215 Cal.App.3d 1517, 264 Cal.Rptr. 740, 751-52 (1989). These cases, however, involve stockbrokers who served as financial advisors or as agents of their clients. Eaton Vance does not argue that this relationship exists when an underwriter simply sells securities to a sophisticated investor at arm’s length. Indeed, that is the position Eaton Vance has abandoned with respect to federal law.
. While FICAL insists (in connection with other issues raised in this lawsuit) that its role was merely that of a stakeholder, FICAL was obligated to protect the bondholders to the extent that the obligation to protect was written into the Trust Indenture. The surety bond provision — • which was indisputably intended for the bondholders' protection — is set forth in the Indenture as a duty of the trustee. Trust Indenture at § 4.06 ("No amount of the initial deposit of Bond proceeds into the Construction Fund shall be released by the Trustee from the Construction *1491Fund unless and until there shall be full compliance with each of the following conditions: .... (2) The Trustee shall have received the Surety Bond_") (emphasis added).
. In the proceedings below, Eaton Vance argued (in response to FICAL) that under California law, as opposed to the New York law relied upon in Meckel, an indenture trustee is not limited to duties set forth within the four comers of the indenture. In support of this proposition Eaton Vance cited Strauss v. Superior Court, 36 Cal.2d 396, 224 P.2d 726 (1950). The Strauss court, however, explicitly left open the possibility that indenture trustees might be able to limit themselves to the stated terms of the trust document by including "exculpatory clauses." Id. at 402, 224 P.2d 726. The Indenture in this case contains such a clause.
. Although we rely on a tort theory of duty, we apply a contractual measure of damages because Bear Steams’ undertaking was of a specific kind: to assist the trustee in carrying out certain of its contractual obligations to the bondholders. A contractual limitation applies because the undertaking was limited to partial performance of a contract.
. Bear Steams goes further and argues that as a matter of causation, its liability for this breach is zero. Bear Steams points out that, according to the OS, the surety bond was meant "[a]s support for the Guaranty.” Because the guaranty expired upon default and acceleration, Bear Steams argues, Eaton Vance would have gotten very little good out of a surety bond that remained alive for another year and a half; a bond that is derivative of a dead guaranty is a dead bond.
But while this seems sound enough as a matter of logic, the fact is that an initial payment of $485,000 was made under the surety bond on December 1, 1988—after default and the expiration of the guaranty. It is at least arguable that another payment would have been due and would have been made. (We emphasize, however, that because obligations under the surety bond are not at issue here, we do not purport to make any determinations with respect to them.)
Bear Steams also argues that the fact that it authorized a surety bond with a premature expiration date does not matter because under § 3.9(c) of the Loan Agreement the developer had the option to provide an "Alternate Credit Facility," which might also have had an earlier expiration date. But this does not mean that Bear Steams' approval of a surety bond with a premature expiration date was permissible. In this case, an Alternate Credit Facility was not provided; rather, a surety bond was provided, and it did not meet the specifications of the Loan Agreement. Section 3.9(c) makes it clear that *1492the developer had the option to provide an Alternate Credit Facility in lieu of the “previously held Surety Bond,” which would then be "promptly surrender[ed].” But here, since no such exchange took place, the terms of the surety bond—which Bear Stearns approved—were dis-positive of the bondholders’ ability to recover, and there can be no dispute that the terms of the surety bond did not comport with what was set forth in the Loan Agreement.
. We are aware that in its sixth claim, Eaton Vance seeks damages based on breach of fiduciary duty and on nondisclosure, while we hold here that it may be entitled to damages based on a non-fiduciary duty and on actions Bear Steams took in assisting the trustee. Thus the theory under which we approve potential relief for Eaton Vance is an imperfect fit with the theory represented in Eaton Vance's own pleadings. Eaton Vance is entitled to rely, however, on the catch-all clause contained in the penultimate sentence of its complaint, in which it states that it seeks "such other and further relief as the Court deems just and proper.”
We do not share Judge Norris’s concern that an award of damages is inappropriate because Eaton Vance seeks only rescission with respect to its seventh claim. The sixth and the seventh claim are based on factually overlapping arguments concerning fiduciary duty, and Eaton Vance does seek damages under its sixth claim. In any event, the catch-all clause cures any variance between the relief requested and the relief which may ultimately be given.
On a broader level, we are not concerned, as Judge Norris is, that the authorities we rely upon were not cited by Eaton Vance, nor even that the theory we rely on does not correspond precisely to any theory advanced by Eaton Vance. We are not restricted to the particular arguments made by the parties. This case illustrates the wisdom of that rule. Eaton Vance asks for too much, and Bear Stearns concedes too little. The truth lies somewhere in between. That's the adversary process at work.
. We have received FICAL’s "Objection of Consideration of Issues and Arguments First Raised in Plaintiffs’ Reply Brief” as well as appellants' response to that objection. We construe the "Objection” as a motion to strike, and we deny that motion, since FICAL has, in the Objection itself, effectively sur-replied to any new issues raised by appellants.
. At oral argument, FICAL contended that the parties’ course of dealing established that “land and building" was a reasonably detailed description, given the limited role FICAL was to play under the Trust Indenture. We have serious doubts about this argument, particularly in light of the fact that when the requisition certificates in question were submitted, the land and the building which stood on it had already been purchased. In any event, FICAL's course of *1495dealing argument at best creates a dispute of fact—it does not lay one to rest.
. This taxonomy, developed in tort cases, appears to apply to contract cases as well. See Bruckman v. Parliament Escrow Corporation, 190 Cal.App.3d 1051, 235 Cal.Rptr. 813, 820 (1987) ("substantial factor” analysis applies in contract cases).
. Mitchell rejected the “but for” jury instruction which had been used in California negligence cases, BAJI 3.75. The court disapproved of the instruction because it required that a defendant's conduct lead "in natural and continuous sequence” to plaintiff's injury. The court thought that this phrase would lead jurors "to focus improperly on the cause that is spatially or temporally closest to the harm.” 1 Cal.Rptr.2d at 919, 819 P.2d at 877. The court clearly did not, however, disapprove of the concept of "but for” causation. Id. at 921 n. 10, 819 P.2d at 879 n. 10 (inviting the Committee on Standard Jury Instructions to draft a new "but for” instruction); Brookhouser, 13 Cal.Rptr.2d at 665. Nor did the Mitchell court suggest that the "but for” concept establishes a threshold for cause in fact imper-missibly lower than that embodied in the "substantial factor" test. Rather, it reasoned that in most cases the substantial factor test and a (properly formulated) "but for” test would bring about identical results. Id., 1 Cal.Rptr.2d at 920, 819 P.2d at 878.
. In order to draw the inferences in favor of the Bondholders with respect to the question of causation — as we must do on summary judgment— we assume that the Bondholders could prevail on the question of whether FICAL breached its duty by disbursing funds on the basis of requisition certificates which lacked the required level of detail.
. The absence of evidence on this point is not surprising. It is difficult to imagine what form the evidence would take. The most direct evidence would presumably be Deringer's deposition, in which he would testify that he either would or would not have diverted funds if he had been compelled to supply more detailed requisition certificates. Given that such testimony could have exposed Deringer to criminal liability, we are not surprised that the Bondholders did not obtain it. Another form of evidence might come from experts on bank fraud, who might be able to testify about the measures taken by banks to police funds such as FICAL's Construction Fund, and on the success rate such measures enjoy.
. The Bondholders, relying on Whitfield v. Lin-demann, 853 F.2d 1298, 1304-05 (5th Cir.1988), cert. denied, 490 U.S. 1089, 109 S.Ct. 2428, 104 L.Ed.2d 986 (1989), argue that the burden of disproving proximate cause rests on the trustee once the beneficiaries have shown a breach. Whitfield, however, noted that there is a split of authority on this question, and the Bondholders have not cited any case in which a California court shifted the burden to the defendant as an initial matter. In the absence of any such authority, the burden of proving causation, at least initially, rests here, as in any other tort or contract case, on the plaintiffs. See Bruckman, 235 Cal.Rptr. at 820.
. Of course, the Bondholders still bear the burden of showing true proximate cause — which in a contract case means showing that their injury "in the ordinary course of things would be likely to result” from the breach. Cal.Civil Code § 3300.
. FICAL also refers to the "analogous” context of cases decided under Cal.Commercial Code § 3420, and cites one of those cases for the rule that “ ‘a bank may escape liability for honoring a check on a faulty or improper endorsement, or even without an endorsement, if the bank can prove that the intended person received the proceeds of the check.’ " Southern Cal. Permanente Medical Group v. Bozinovski, 148 Cal.App.3d 503, 196 Cal.Rptr. 150, 154 (1983) (quoting Northwest Bank of Clearwater v. Bentley, 413 So.2d 480, 481 (Fla.App.1982)). Bozinovski, however, involved a once-only transaction rather than an ongoing series of withdrawals, and hence has litde relevance to this case, where the series of transactions provides the basis for the Bondholders' deterrence theory.
. The Bondholders question the credibility of Hall and Pochos, noting that at the time the witnesses testified, both of their law firms were defendants in this litigation. The Bondholders do not explain, however, in what way it would benefit either of the firms to establish that the surety approval clause was a typographical error. Indeed, since both firms were co-defendants of FICAL, the purported bias in at least one sense should have run in the other direction: the lawyers should have wanted the clause declared valid, so that FICAL’s fault would be greater and their firms' fault (and liability) proportionately lesser.
. The Bondholders’ argument that reformation is inappropriate because it impinges on their rights as third party beneficiaries fails for a similar reason. Cal.Civil Code § 3399 states that reformation is appropriate only "[in] so far as it can be done without prejudice to rights acquired by third persons, in good faith and for value.” Since the Bondholders do not purport even to have known about the surety approval clause at the time they made their purchases, they cannot be said to have given value for any rights accruing to them under that clause.
. M.B.A.F.B. Fed. Credit Union v. Cumis Ins. Soc'y, Inc., 681 F.2d 930, 932-33 & dissent n. 1 (4th Cir. 1982), cited by the Bondholders, is distinguishable. In that case the Fourth Circuit held that it was not error for the trial court to have admitted a conclusive statement made by a witness ("I know [defendant] was aware of the purchase price of the properties”), where the foundation for the statement was not entirely clear ("I don't recall [discussing the matter with defendant]"). Here, by contrast, there is no conclusive statement at all. There is simply the testimony that the witnesses do not remember one way or the other whether they had the conversations with Holmstedt.
. They did argue that Holmstedt did not hand deliver the letter to Sulzberger, but that is not the same as saying that Sulzberger never received it.
. The law firms argue that there is no per se requirement that the district court make express findings of fact. The problem with the district court's order, however, is that it is so ambiguous (if the law firms’ interpretation is given credence) that a reviewing court cannot tell whether any fact-finding was done. Under California law a good faith settlement determination is clearly a factual matter, Owen v. United States, 713 F.2d 1461, 1466 (9th Cir. 1983); here, more specificity is necessary in order to determine whether the court made any factual determination at all.
. While FICAL explicitly makes this argument only with respect to Jones Hall, and only in connection with the surety approval clause, the references FICAL makes to Brobeck in connec-tíon with the reasonable detail requirement and the Loan Agreement suggest that it ascribes a similar role to both of the law firms.