Opinion for the Court filed by Circuit Judge WALD.
WALD, Circuit Judge:In this consolidated case, two parties to a natural gas transportation contract petition for review of a rate-setting order of the Federal Energy Regulatory Commission (“FERC” or “Commission”). Western Resources, Inc. (“Western”),1 a Kansas intrastate pipeline, contracted with Panhandle Eastern Pipe Line Company (“Panhandle”), a large interstate pipeline, for “forward haul” transportation of gas during the summer off-peak season from Kansas to Michigan storage sites and “backhaul” return of the gas to Kansas during the peak winter season under Rate Schedule T-53. Western seeks review of the Commission’s order approving an increase in the forward-haul rate for off-season transportation. Western argues that FERC’s decision runs counter to its regulations, fails to recognize the significant systemic advantages accorded by off-peak forward-haul service and the limitations of the parties’ contract, and deviates from the parties’ agreement specifying that transportation rates should increase commensurately with sales rates. Panhandle seeks review of the Commission’s order requiring backhaul rates to be set at one-half of forward-haul rates in order to reflect the seasonal benefits of decreased capacity constraints. Panhandle contends that the Commission failed to bear its burden of proof in devising this rate structure, made an unwarranted departure from the terms of the parties’ settlement agreement, breached its own regulations, and impermissibly extended its order retroactively. We remand the Commission’s order as to the forward-haul rate so that the Commission may more fully explain its decision, remand as to the backhaul rate so that the Commission may reexamine whether it wishes to proceed under § 4 or § 5 of the Natural Gas Act (“NGA”), 15 U.S.C. §§ 717c(e), 717d(a) (1988), and remand for reconsideration of the refund order in light of the foregoing.
I. Background
A Factual Context
Western entered into a transportation agreement2 with Panhandle on May 17,1982 for the “forward haul” of natural gas from Reno County, Kansas to storage in Michigan during the off-peak summer months and the “backhaul” return of gas to Reno County, Kansas for sale during the peak winter months. The contract included a strict seasonal limitation, permitting forward hauls only in the summer off-peak months, fixed points of receipt, and firm, nondiseountable rates. Since gas flows on Panhandle’s system from west to east, the forward-haul service required the introduction of gas into the system at the upstream point, Reno County, and the removal of gas at the downstream point in Michigan. The backhaul service, in contrast, involved the removal of gas at the upstream point and the introduction of gas at the downstream point3 Thus, the backhaul *1571service, but not the forward-haul service, involved a reduction in the amount of gas flowing between the upstream point and the downstream point. Because of the contract’s seasonal limitations, forward hauls only occurred during the period in which the pipeline was used least, and backhauls, which alleviated the burden on the system between the upstream and downstream points, only occurred when the capacity of the pipeline was most constrained. In recognition of the distinctions between these forms of transportation, the parties’ contract employed two different methodologies in calculating rates. The forward-haul rate for Rate Schedule T-53 service was based on Panhandle’s system unit cost per 100 miles, which worked out to approximately 20.05 cents per Mcf,4 and the backhaul rate was set at a fixed one cent per Mcf. See Joint Appendix (“J.A”) at 104-05.
B. Procedural History
On August 31, 1987, Panhandle filed revised tariff sheets in Docket No. RP87-103-000. Among other adjustments, the new tariff sheets proposed to change Rate Schedule T-53 by requiring a full allocation of costs for all services, thus increasing the rates for backhaul and forward-haul services to a level equivalent to that for forward-haul open access transportation service under Rate Schedule PT.5 On September 30, 1987, the Commission issued an order accepting for filing and suspending certain of the tariff sheets, with the result that the increased T-53 rates became effective March 1,1988 subject to refund. See Panhandle Eastern Pipe Line Co., 40 F.E.R.C. ¶ 61,369, reh’g granted in part, 41 F.E.R.C. ¶ 61,170 (1987). On September 30, 1988, Panhandle filed another rate increase petition with the Commission pursuant to § 4 of the NGA in Docket No. RP88-262-000. These rates became effective on April 1, 1989. See Panhandle Eastern Pipe Line Co., 45 F.E.R.C. ¶ 61,145 (1988), reh’g granted in part, 46 F.E.R.C. ¶ 61,268 (1989). As a result, the rates at issue in this case apply only to a thirteen-month “locked-in” period from March 1,1988 to March 31, 1989.
The Commission held a hearing between March 17 and April 29,1988 on the propriety of the locked-in rate hike. On March 22, 1990, the parties filed a settlement agreement with the Commission that sought to resolve several of the issues litigated in the hearing. Simultaneously, the parties filed a Stipulation and Agreement in the superseding rate proceeding, Docket No. RP88-262-000, with the expectation that each settlement offer would be contingent on the other. On May 22, 1990, the presiding Administrative Law Judge (“ALJ”) declined to certify the settlement in Docket No. RP88-262-000. The parties agreed on June 26,1990 to allow certification of the settlement in this docket while continuing litigation in Docket No. RP88-262-000. The settlement, approved by the Commission on November 26, 1991, reserved one issue for the Commission’s disposition in Article III:
Whether the rates for service to [Western] under Rate Schedule T-53[,] which is performed pursuant to section 7 authorization[,] which predate the issuance of Order 436 shall be equivalent to the maximum rates for services provided under Rate Schedule PT.
Panhandle Eastern Pipe Line Co., 57 F.E.R.C. ¶ 61,265, ¶ 61,858 (1991). Article III § 5 set forth the procedures by which the parties agreed to litigate the reserved issue, specifically contemplating the contingency that the Commission might determine that Western was entitled to a refund. See J.A. at 296.
In its initial order, FERC approved the forward-haul rate increase as “entirely consistent with the Commission’s regulations,” but determined that “a reasonable maximum charge for the backhaul portion of the Rate Schedule T-53 service is a rate equal to one half the forward haul rate[,] because a back-*1572haul service benefits Panhandle by creating capacity that may be used by other shippers.” 57 F.E.R.C. 161,265, ¶ 61,863. The Commission ordered Panhandle to refund monies paid by Western in excess of the adjusted backhaul rate, “as provided for in article III of the settlement.” Id. Both Western and Panhandle filed timely requests for rehearing, Western urging that the Commission decrease the forward-haul rate, and Panhandle requesting that the Commission reverse its determination with respect to the adjusted backhaul rate.
The Commission rejected both petitions for rehearing. With respect to the forward-haul rate, the Commission affirmed its order approving the rate increase. The Commission concluded that the settlement agreement did not require that it consider whether Rate Schedule T-53 service should be equivalent to sales rates under Rate Schedule TS-4. The Commission specifically denied Western’s contention that the forward-haul rate should be decreased in order to maximize throughput during the off-peak summer months. Noting that the reserved issue focused on a comparison between Rate Schedule T-53 and Rate Schedule PT, the Commission observed that Rate Schedule PT did not differentiate between summer and winter rates. Accordingly, the Commission concluded that Western “has only in effect suggested that Rate Schedule PT service provided during the summer should have lower rates.” 59 F.E.R.C. ¶ 61,245, ¶ 61,874 (1992) (emphasis added). Responding to Western’s argument that the inferiority of its limited service required lower rates, the Commission determined that “[a] solution to this problem may be to allow [Western] more flexibility under Rate Schedule T-53.” Id. Although the Commission conceded that reduced reservation charges might be appropriate for summer service, it declined to mandate a reduction, because “no party, including [Western], has provided evidence which would help the Commission determine with any certainty what those reservation charges should be.” Id.
As to the backhaul question, the Commission relied on its Opinion No. 369, 57 F.E.R.C. ¶ 61,264 (1991), issued contemporaneously with the initial order, in which it reasoned that backhauls like that provided under Rate Schedule T-53 create usable capacity for the pipeline “when and where that capacity is needed.” 59 F.E.R.C. at ¶ 61,873 (citing 57 F.E.R.C. ¶ 61,264, 161,839-40). The Commission affirmed its order that the backhaul rate should be set at one-half the forward-haul rate in order to reflect these systemic benefits, stating that the figure “gives a reasonable approximation of the benefits conferred and is appropriate until a better method can be established.” Id.
Western and Panhandle petitioned this court for review under § 19(b) of the NGA, 15 U.S.C. § 717r(b). We now grant both petitions and remand to the Commission for further proceedings consistent with this opinion.
II. Analysis
A. Standard of Review
As this court stated in Columbia Gas Transmission Corp. v. FERC, 628 F.2d 578 (D.C.Cir.1979), “[j]udicial scrutiny under the [NGA] is limited to assuring that the Commission’s decisionmaking is reasoned, principled, and based upon the record.” Id. at 593. Ordinarily, this court is “without authority to set aside any rate selected by the Commission which is within a zone of reasonableness.” Public Serv. Comm’n v. FERC, 813 F.2d 448, 451 (D.C.Cir.1987) (citing Permian Basin Area Rate Cases, 390 U.S. 747, 767, 88 S.Ct. 1344, 1360, 20 L.Ed.2d 312 (1968)). Nonetheless, our review must ensure that the Commission has acted consistently with the statutory framework, both in adhering to the statutory allocation of burdens of proof and in “demonstrating the connection between the facts found and the choice made.” ANR Pipeline Co. v. FERC, 771 F.2d 507, 516 (D.C.Cir.1985). To this task we now turn.
B. The Forward-Haul Rate
Western makes a variety of arguments in support of its contention that the Commission erred in approving Panhandle’s proposed forward-haul rate.- First, Western maintains that the forward-haul rate did not *1573reflect either the cost or the quality of the service provided. Because the forward-haul service under Rate Schedule T-53 was non-discountable, limited to the off-peak season, and lacking in receipt-point flexibility, Western urges that it was markedly inferior to the PT rate. Moreover, insofar as the service provided the pipeline with extra throughput — and, consequently, needed revenue recovery — during traditional periods of “down time,” it conferred benefits on the pipeline that should have been recognized in more favorable rates. Western contends that existing FERC regulations required the Commission to set rates for off-peak periods that reflected seasonal variations in cost and provided incentives to maximize throughput. Second, Western argues that the pipeline’s new rate arbitrarily discriminated against customers who did not purchase gas from Panhandle. Western points to the fact that Panhandle set the price for off-peak forward hauls made in connection with its sales contracts to cover only minimal administrative costs. Western contends that this discrimination violated the Commission’s Order No. 636, 130,939, 57 Fed.Reg. 13,267 (1992), which requires pipelines to provide a similar quality of transportation service whether gas is purchased from the pipeline or from another seller. Third, Western argues that the contract between Western and Panhandle itr self restricted Panhandle’s ability to increase the T-53 service rate without similarly increasing its sales rates. We consider the first argument separately and the second and third arguments in tandem.
1. Cost and Quality of Service
Service under the Rate Schedule T-53 contract between Western and Panhandle differed in several respects from service under Rate Schedule PT. In contrast to an open access contract, which includes no seasonal constraint, the parties’ agreement strictly limited the season in which Western could inject gas into the system for forward haul to the period between April 1 and October 31, months during which the pipeline’s flow traditionally is at an ebb. J.A at 67. The contract permitted Western to engage in winter withdrawals of gas via backhauls only between November 1 and March 31, a period in which the pipeline operates at near-maximum capacity. Id. The contract limited the sites of gas delivery to specific points of interconnection in Kansas and the storage state. Id. at 67-68. Finally, the contract’s rate schedule provided for no discountability.6 Id. at 72-73. Western argues that, in view of the marked differences between Rate Schedule T-53 service and Rate Schedule PT service, the Commission erred in upholding equivalent rates for the T-53 forward haul.
Moreover, Western argues, because of the seasonal limitations imposed on Western’s service, the contract conferred appreciable advantages on the Panhandle system and failed to exact certain costs associated with peak-period transportation. First and foremost, the winter backhaul service afforded the pipeline more capacity to transport gas between the upstream point, at which gas was removed, and the downstream point, at which gas was reinserted. In fact, the Commission recognized the possibility of just these benefits in the context of mandating a lower backhaul rate, discussed infra. Second, the summer forward-haul service allegedly enabled the pipeline to recover costs during the period in which the pipeline was used least. The forward-haul rate for off-peak service, maintains Western, properly should recognize the fact that few customers would otherwise seek out the pipeline’s services during this period. Other things being equal, Western asserts, the pipeline benefit-ted tangibly from the increased throughput stimulated by the erstwhile favorable Rate Schedule T-53. Finally, as a summer forward-haul customer, Western faced no competition for scarce pipeline capacity. During the winter, when customers abound, higher rates both offset the costs of greater burdens on the system and serve as a rationing mechanism. Paradoxically, Western suggests, the Commission took seasonal factors into account in decreasing the backhaul rate, but *1574refused to take cognizance of them in its consideration of the forward-haul rate.
Western bolsters its argument by several of the Commission’s regulations that touch upon seasonal rates. At 18 C.F.R. § 284.-7(c)(2), the Commission sets forth objectives for rate-setting during off-peak periods, stating that:
Rates for firm service during off-peak periods and for interruptible service during all periods should maximize throughput.
Moreover, 18 C.F.R. § 284.7(d)(3) provides that:
Any rate filed for service subject to this section must reasonably reflect any material variation in the cost of providing the service due to:
(i) Whether the service is provided during a peak or an off-peak period.
Finally, as Western notes, 18 C.F.R. § 284.-7(d)(4)(i) provides that:
Any maximum rate filed under this section must be designed to recover on a unit basis, solely those costs which are properly allocated to the service to which the rate applies.
Although these regulations appear to leave considerable discretion to the rate-setter, they nevertheless clearly contemplate that different costs may be attributable to different services in different seasons. Western urges this court to find that the Commission violated these regulations by approving an off-peak forward-haul rate that, like the PT rate, included a full allocation of capacity costs, when the T-53 service imposed no such capacity costs on the system.
Western’s arguments are not insubstantial.7 We begin by recognizing that, because it was the source of the proposed change, Panhandle bore the burden of persuasion under § 4 of the NGA to justify a departure from the presumptively just and reasonable preexisting rate structure, under which the forward-haul rate for T-53 service was significantly lower than that for Rate Schedule PT service. See 15 U.S.C. § 717c(e) (1988); East Tennessee Natural Gas Co. v. FERC, 863 F.2d 932, 937 (D.C.Cir.1988). Section 4 limits the Commission to two courses of action, “acceptance (in whole or in part) or rejection of the pipeline’s proposed rates.” Sea Robin Pipeline Co. v. FERC, 795 F.2d 182, 183 (D.C.Cir.1986). Under its § 4 authority, the Commission need only articulate the bases for its determination that the proposed rate is just and reasonable, demonstrating a sound and rational nexus between the facts and its conclusions.8
Even under such a relaxed standard, we find that the Commission’s opinion falls short in making manifest the requisite connection. Somewhat perplexingly, the Commission appeared to accept (or at least, not to reject) many of Western’s substantive arguments even as it approved the increased forward-haul rate over Western’s protest. In its Order on Rehearing, the Commission conceded that seasonal rate differences might be necessary in order to ration rights to winter capacity, but observed that “no party, including [Western], has provided evidence which would help the Commission determine with any certainty what those [adjusted rates] should be.” 59 F.E.R.C. at ¶ 61,874. The Commission acknowledged that Rate Schedule T-53 service lacked the flexibility of PT service, but determined that this factor suggested that Rate Schedule T-53 service should be adjusted for more flexibility, not priced more cheaply.9 The Com*1575mission concluded, “[i]n any event, [Western] has not shown what value should be attached to this feature of Rate Schedule PT service so that the Commission could determine an appropriate rate adjustment.” Id, at ¶ 61,-875. Although the Commission apparently found some measure of logic to Western’s arguments, it declined the opportunity to interpose its own rate under § 5 of the NGA. In approving the new rate notwithstanding these lingering questions, the Commission implicitly found the proposed increase just and reasonable.
To be sure, we approach this finding with the deference befitting a reviewing court’s inquiry into rate-setting. The Commission, possessed as it is of considerable expertise, is entitled to hold sway provided that its conclusion “is supported by substantial evidence and reached by reasoned decisionmaking— that is, a process demonstrating the connection between the facts found and the choice made.” ANR Pipeline Co. v. FERC, 771 F.2d 507, 516 (D.C.Cir.1985) (citing Burlington Truck Lines, Inc. v. United States, 371 U.S. 156, 167-68, 83 S.Ct. 239, 245-46, 9 L.Ed.2d 207 (1962)). Taking these principles firmly in hand, we nonetheless conclude that the Commission’s rationale fails to dispel certain vexing questions about the propriety of assessing equivalent rates for two very different services, the off-peak forward haul and the unlimited, open access transportation. Particularly given the Commission’s apparent concession that seasonal disparity in rates might be warranted in this case and the Commission’s rate objectives embraced in regulations, we are unpersuaded that the Commission’s decision approving the increased forward-haul rate comports with reason and logic.
In its brief, the Commission makes several arguments that intensify our discomfort. The Commission argues for the first time that it viewed the forward- and backhaul rates “as parts of a closely interrelated service,” and that it “attempted to give recognition to the capacity advantages that the T-53 service provides overall to Panhandle by requiring that the backhaul rate be halved_” Brief for Respondent at 39. Despite this concession on brief, the orders below gave no indication that the Commission intended the adjusted backhaul rate to reflect supposed benefits of the off-peak forward-haul service. Instead, the Commission repeatedly lamented that Western had provided it no means of quantifying any possible benefits of the off-peak forward haul, and that it therefore had to accept Panhandle’s proposed rate hike at face value. See 59 F.E.R.C. at ¶ 61,874. Obviously, a new gambit undertaken by the Commission in its brief at this stage cannot be used to rationalize the Commission’s action below, due to our longstanding rule that “courts may not accept appellate counsel’s post hoc rationalizations for agency action.” Algonquin Gas Transmission Co. v. FERC, 948 F.2d 1305, 1316 (D.C.Cir.1991) (citing Motor Vehicle Mfrs. Ass’n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 50, 103 S.Ct. 2856, 2870, 77 L.Ed.2d 443 (1983)). By the same token, we also refrain from1 drawing the negative inference from new arguments at this stage that the Commission intends to abandon or undercut rationales employed below. Nonetheless, we do observe that this new argument seems further to underscore concerns already discussed about the Commission’s apparent agreement with Western’s underlying premise that off-peak forward-haul service differs from Rate Schedule PT service in ways that might justly and reasonably be reflected in lower rates — whether in the forward-haul rate itself, as Western urges, or in the back-haul rate, which the Commission’s counsel now asserts for the first time motivated the Commission’s 50% cut. Yet the fact remains that the Commission’s forward-haul decision itself indicates outright rejection of seasonal considerations.
The Commission also contends in its brief that Rate Schedule T-53 incorporated certain benefits lacking in PT service that compensated for its other limitations. Specifically, the Commission now argues that Western’s service “could not be abandoned without the exercise of the Commission’s express abandonment authority under section 7(b) of the *1576NGA, ... whereas open-access transportation under Rate Schedule PT was subject to ‘pre-granted abandonment.’ ” Brief for Respondent at 40. The Commission urges now — although it did not so conclude below— that because of the abandonment factor, PT service is not necessarily superior to T-53 service. Although the issue of abandonment might cast the rate-setting issue in a different light, we are unwilling to deviate from the time-honored rule that a reviewing court “must judge the propriety of [agency] action solely by the grounds invoked by the agency.” SEC v. Chenery Corp., 332 U.S. 194, 196, 67 S.Ct. 1575, 1577, 91 L.Ed. 1995 (1947). Accordingly, we take no stock in the Commission’s post hoc justification. We observe, however, that again, by offering a new factor at this late stage, the Commission reinforces nagging doubts that, without more, seasonal limitations should have translated into lower rates.
We remand to the Commission for reconsideration of the T-53 forward-haul rate in light of regulations that appear to endorse seasonal distinctions, the Commission’s apparent recognition below (repeated in its brief) that the off-peak forward-haul rate differs materially from the PT rate, and our belief that the Commission cannot adequately justify at 224% increase in the off-peak forward-haul rate structure merely by asserting that the affected customer provided no substitute. Even when the Commission does not itself bear the burden of proof, the Commission’s responsibility to ensure that all approved rates are just and reasonable requires more explanation than it gave here for a remarkable deviation, from the presumptively just and reasonable status quo.
2. Parity With Sales Rates
Western also argues that Panhandle’s forward-haul rate hike discriminates against nonsales customers, who pay only a modest administrative fee for off-peak forward hauls made in conjunction with the sale of gas. In approving the rate increase, Western maintains, the Commission countenanced discrimination between the two services in violation of its regulations and the parties’ original agreement. Without ruling on the substance of Western’s arguments, the Commission concluded that the comparison of sales and transportation rates was beyond the purview of the reserved settlement issue. As a threshold matter, therefore, we must determine whether the Commission reasonably interpreted the settlement agreement to preclude examination of Western’s claims. We conclude that the Commission acted reasonably in light of the terms of the settlement and the circumstances surrounding the litigation. Accordingly, we do not venture beyond the threshold to consider the merits of Western’s argument.
In National Fuel Gas Supply Corp. v. FERC, 811 F.2d 1563 (D.C.Cir.1987), this court determined that the Commission’s reading of a settlement agreement is entitled to deference even when the issue involves the proper construction of the agreement’s terms.10 The inquiry, the court reasoned, should proceed along the lines of the now-familiar Chevron analysis: “if the intent of the parties on the particular issue is clearly expressed in the document, ‘that is the end of the matter.’” Id. at 1572 (citing Chevron U.S.A Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 842, 104 S.Ct. 2778, 2781, 81 L.Ed.2d 694 (1984)). If the parties’ agreement does not speak directly to the problem, however, we are bound to uphold a reasonable agency interpretation of the agreement. To be sure, we need not accept “an agency interpretation that black means white. However, if the choice lies between dark grey and light grey, the conclusion of the agency, unmistakably possessed as it is of special expertise, in favor of one or the other will have great weight.” Id. (citing Consolidated Gas Supply Carp. v. FERC, 745 F.2d 281, 291 (4th Cir.1984) (citation omitted)).
Article III of the settlement agreement specifically reserves for the Commission the issue “[w]hether the rates for service to *1577[Western] under Rate Schedule T-53[,] which is performed pursuant to section 7 authorization[,] which predate the issuance of Order 436 shall be equivalent to the maximum rates for services provided under Rate Schedule PT.” 57 F.E.R.C. at ¶ 61,858. The reserved issue does not address sales rates, which seem to present altogether different issues. The Commission interpreted the reserved question to focus only on the comparison between Rate Schedule PT and Rate Schedule T-53, not to extend to consideration of sales rates. As the Commission observed, Western and Panhandle could revisit the issue of Rate Schedule TS-4 sales rates and their effect on Rate Schedule T-53 in Panhandle’s more recent rate case in Docket No. RP91-229-000. In light of the language of the settlement issue, which makes no reference to sales rates, we conclude that the Commission was well within its discretion in determining that consideration of sales rates was beyond the scope of the agreement. Accordingly, we decline to second-guess the Commission’s decision to pass on that issue.
C. The Backhaul Rate
In its decision below, the Commission accepted Western’s argument that winter back-haul service allowed Panhandle to transport more gas during periods of strained capacity between the upstream point and the downstream point. The Commission agreed that these benefits should translate into lower rates, concluding that “the appropriate Rate Schedule T-53 backhaul rate is one half the Rate Schedule T-53 forward haul rate_” 59 F.E.R.C. at ¶ 61,873. The Commission determined that this figure, proposed by Western, “gives a reasonable approximation of the benefits conferred and is appropriate until a better method can be established.” Id.
Panhandle challenges this conclusion on several grounds. First, Panhandle claims that the separate issue of backhaul rates was not properly before the Commission, because the parties’ settlement agreement restricted FERC to consideration of whether T-53 rates could be set equal to PT rates, not whether backhaul rates and forward-haul rates should be set differently. Second, Panhandle argues that the backhaul rate ultimately approved by the Commission was the Commission’s own creation, a new figure falling under the Commission’s § 5 authority under the NGA. Consequently, Panhandle contends, the Commission was required, but failed, to bear the burden of proof on the basis of substantial evidence that its proffered new rate was “just and reasonable.” Finally, Panhandle claims that under § 5, FERC was not authorized to grant refunds. We consider these issues in turn.
1. Scope of the Reserved Question
The settlement agreement reserved the single question whether T-53 rates should be set equal to rates under Rate Schedule PT. Panhandle contends that this issue is narrow, entailing a literal comparison of T-53 and PT rates without distinction between forward-haul and backhaul charges. The Commission rejoins that its decision answers the question affirmatively for forward-haul rates and negatively for backhaul rates, and is entirely consistent with the reserved question. To evaluate this claim, we hearken back to National Fuel Gas Supply Corp. v. FERC, 811 F.2d 1563, 1569 (D.C.Cir.1987), in which this court underscored the principle of deference to the agency’s construction of settlement agreements. Although it is true that the reserved question is not parsed into forward-haul and backhaul components, we agree with the Commission that the issue is ambiguous. Forward-haul and backhaul rates were historically set much differently by the parties, and proper resolution of the reserved question seemingly requires consideration of the two components separately. In light of these factors, we find the Commission’s conclusion reasonable and worthy of deference. As the Commission notes, the difference between the forward-haul and backhaul rates was at the core of the parties’ disagreement.
2. The Commission’s Authority under the NGA
The remaining two issues raised by Panhandle — whether the Commission sustained its burden of proof and whether the Commission was authorized to apply the ad*1578justed back-haul rate retroactively — require characterization of FERC’s action in approving a backhaul rate equal to one-half of the approved forward-haul rate. Under the NGA, the allocation of the burden of proof and the propriety of ordering refunds depend on the source of the proposed rate change. See East Tennessee Natural Gas Co. v. FERC, 863 F.2d 932, 937 (D.C.Cir.1988); Tennessee Gas Pipeline Co. v. FERC, 860 F.2d 446, 449 (D.C.Cir.1988). Section 4, which governs the approval or rejection of rates proposed by the pipeline, mandates that the pipeline bear the burden of proof that its proposed rate is just and reasonable and permits refunds in certain narrowly drawn circumstances.11 See 16 U.S.C. § 717c(e) (1988). Section 5 governs situations in which the Commission imposes rates of its own creation or at the behest of a third party. This section requires FERC to bear the burden of proof that its proffered rate is just and reasonable and bars retroactivity altogether. See 15 U.S.C. § 717d(a) (1988). Panhandle argues that the Commission’s determination of the backhaul rate in the proceedings below ventured beyond the exercise of its § 4 authority and into the realm of § 5. As a result, Panhandle argues, the Commission was required by statute to bear a burden that it did not meet in this instance.
The Commission responds by contending that its action fell squarely within the ambit of its § 4 powers. The Commission characterizes its action as a partial acceptance of the pipeline’s proposed rate hike. Under § 4, the pipeline was required to convince the Commission that its proposed rate .was just and reasonable. FERC maintains that it found that the pipeline had sustained this burden, but only to the extent that the new backhaul rate did not exceed one-half of the forward-haul rate. Because § 4 contemplates acceptance in whole or in part of pipeline proposals, see Sea Robin Pipeline Co. v. FERC, 795 F.2d 182, 183 (D.C.Cir.1986), the Commission urges us to find that the pipeline — and the pipeline only — bore the full burden of proof in connection with the new rate, and that it sustained that burden only by half.
This court “has consistently disallowed attempts to blur the line between §§ 4 and 5.” Public Serv. Comm’n v. FERC, 866 F.2d 487, 491 (D.C.Cir.1989). As we complained four years ago, “[o]n four occasions in the last three years this court has reviewed Commission efforts to compromise § 5’s limits on its power to revise rates. On each the court has repelled the Commission’s gambit. This is number five.” Id. at 488-89. We now make it an even six. The approved backhaul rate at issue here is methodologically distinct from the one proposed by the pipeline. Panhandle had proposed uniform forward-haul and backhaul rates that would recover a full allocation of its fixed costs each way. See J.A. at 399. Panhandle argued that, because backhaul service presupposed working forward-haul service, the services should be priced identically. The Commission responded by differentiating between forward-haul and backhaul rates on the grounds that the seasonal backhaul service conferred cognizable benefits on Panhandle’s system that necessitated alteration of the rate structure. The Commission set the backhaul rate at one-half the forward-haul rate, and in so doing reached beyond approval or rejection of the pipeline’s proposal to adoption of an entirely different rate design.
The Commission attempts to distinguish other cases in which we insisted upon compliance with § 5 when the Commission imposed a new rate on the grounds that in those cases, the Commission reached out to alter aspects of the rate structure that the pipeline had not proposed to change. Although there *1579are differences, these cases provide ample support for the conclusion that FERC should bear the burden under § 5 whenever it moves beyond rejection of a proposed rate to the task of redesigning it. See, e.g., Tennessee Gas Pipeline Co. v. FERC, 860 F.2d 446, 454 (D.C.Cir.1988) (quoting with approval an ALJ’s statement to this effect); Sea Robin Pipeline Co. v. FERC, 795 F.2d 182, 187 (D.C.Cir.1986) (reasoning that Commission moves out of § 4 range when approved rate methodology deviates from that proposed by the pipeline).
Although we find controlling the methodological distinctions between the proposed and imposed rates so as to mandate compliance with § 5’s strictures, we must also examine the Commission’s principal contention that our precedents allow it to accept proposals “in part.” The Commission argues that under § 4, it may not only approve one prong of a rate proposal and reject another; it may also approve part, but not all, of a single proposed rate. In Sea Robin, we stated that “[s]ection 4 limits the Commission’s authority to acceptance (in whole or in part) or rejection of the pipeline’s proposed rates.... ” 795 F.2d at 183 (emphasis added). Moreover, § 4(e) of the NGA allows for refunds of “the portion of such increased rates or charges by [the Commission’s] decision found not justified.” 15 U.S.C. § 717c(e) (1988) (emphasis added). Some support for the Commission’s proposition thus may be fashioned from snippets of statute and precedent. Of course, our responsibility for construing the NGA requires us to go beyond consideration of mere individual strands, to determine how the various provisions of the Act are interwoven to achieve Congress’ purpose.
After careful consideration of the statutory framework, we cannot accept the Commission’s argument that § 4 permits it to approve any rate, no matter how materially different from that proposed by the pipeline, so long as it can be viewed as a “part” of the original request. We appreciate that minor deviations from the pipeline’s proposed rate based, for example, upon differences as to the extent of specific cost items, may be handled in a § 4 proceeding, but the imposition by the Commission of only half of a proposed rate surely requires more. When the rate imposed by the Commission differs significantly from that proposed by the pipeline, it can no longer be attributed to the pipeline — at least without the pipeline’s consent — so as to qualify for § 4 treatment. Accordingly, we find the Commission’s argument that the 50% backhaul rate was justified as a “partial approval” of the pipeline’s request is precluded by the statutory design, as well as by our own precedent. See, e.g., Sea Robin, 795 F.2d at 187 (“The rate methodology FERC imposed on Sea Robin was not proposed by the pipeline; thus, the order cannot represent an approval, in whole or part, of changes suggested by Sea Robin.”); ANR Pipeline, 771 F.2d at 513.
In sum, we hold that § 4 cannot accommodate the Commission’s action below, in which the pipeline proposed a rate that differed substantially from its old rates,12 and the Commission responded by setting backhaul rates at half of forward-haul rates in order to recognize benefits winter backhauls conferred on the system. Not only did the Commission set a rate different from that proposed by the pipeline, it also employed a completely different strategy in quantifying distinctions between the two kinds of service. If the Commission wished to impose its own rate, the Commission was required to bear the burden of proving that it was just and reasonable in a § 5 proceeding.
3. Application of the § 5 Burden Scheme
Under the NGA, an action may originate as a § 4 proceeding only to be transformed later into a § 5 proceeding. In imposing its own rate under these circumstances, the Commission must make three findings: first, it must conclude under § 4 that the pipeline failed to carry its burden of proof that the proposed rate was just and reasonable; second, it must itself demonstrate that the default position, the prior rate, is no longer just and reasonable; and third, it must establish that its substitute rate is just and reasonable. See Tennessee Gas Pipeline Co. v. FERC, 860 F.2d 446, 456 (D.C.Cir.1988) (FERC *1580must first determine “that the presumptively just and reasonable existing rate is no longer just and reasonable”) (emphasis in original); Sea Robin Pipeline Co. v. FERC, 795 F.2d 182, 184 (D.C.Cir.1986) (FERC must find “that the existing rate is unjust or unreasonable and the proposed new rate is both just and reasonable”); ANR Pipeline Co. v. FERC, 771 F.2d 507, 514 (D.C.Cir.1986) (same). In monitoring whether the Commission has satisfied its burden when it imposes a new rate under § 5, this court has assumed an active stance, requiring that the Commission’s conclusion be “supported by substantial evidence and reached by reasoned deci-sionmaking.” Id. at 516.
The Commission’s order with respect to the backhaul issue in this case is terse, referring for full discussion to companion opinions numbered 369 and 369-A, which pertain to the 1988 rate period. See 57 F.E.R.C. ¶ 61,264, ¶ 61,839-40 (1991); 59 F.E.R.C. ¶ 61,244. The companion opinions, the first issued contemporaneously with the initial order and the second an order on rehearing issued contemporaneously with the Commission’s order on rehearing in this case, reject Panhandle’s proposed increase to a backhaul rate and impose an adjusted rate at Western’s suggestion. The Commission found that peak backhauls confer a substantial benefit on the pipeline by reducing seasonal capacity constraints and saving costs of compression. The Commission concluded that, although Panhandle’s system was not so overburdened as to require capacity rationing, capacity constraints did encumber the system significantly.. In the decision below, the Commission accepted the same adjusted backhaul rate elaborated in the companion opinions, set at one-half of the forward-haul rate, which it termed a “reasonable approximation of the benefits conferred[,] ... appropriate until a better method can be established.” 59 F.E.R.C. at ¶ 61,873.
In drawing this conclusion, the Commission made no effort to demonstrate that the preexisting backhaul rate, which it had previously approved, had somehow become unjust or unreasonable. As stated above, pursuant to § 5, the Commission “must first find that the existing provision is unjust or unreasonable.” ANR Pipeline, 771 F.2d at 514; accord Tennessee Gas Pipeline, 860 F.2d at 456; Sea Robin, 795 F.2d at 187. The Commission’s action setting the backhaul rate lower than the forward-haul rate is more consistent with the preexisting rate structure than with the pipeline’s proposal. Without any findings demonstrating why a departure is warranted, the Commission cannot set out on uncharted territory to impose its own rate.
Panhandle also argues that the Commission failed to establish how the facts warranted the new backhaul rate it actually imposed. The Commission explains its failure to impose a more precise backhaul rate by the fact that neither party offered any means of quantifying the precise benefits involved. We reject this argument, because the Commission, not the other two parties, bore the burden of establishing that its rate was just and reasonable. The Commission’s explanation suggests that the burden lay elsewhere, and that it set this rate as an interim measure until another party bore its burden and recommended a better rate. Because we find that the Commission misperceives its own responsibility in imposing the new back-haul rate, we remand so that the Commission may rechart its course.
The Commission imposed the backhaul rate below while laboring under a misconception about the breadth of its § 4 authority. It is impossible, in the context of § 4, to justify the Commission’s action setting the backhaul rate at one-half of the forward-haul rate. Neither can the Commission justify its action under § 5 on this record; it failed to make the requisite showing below that the preexisting rate had become unjust and unreasonable and that its own rate was just and reasonable. The Commission has therefore acted inconsistently with both § 4 and § 5 of the NGA. We remand so that the Commission can reconsider its two options and strike out on a path that adheres faithfully to the statutory scheme.
4. The Commission’s Refund Order
The Commission ordered Panhandle to refund to Western any money paid under the proposed backhaul rate during the *1581locked-in period in excess of one-half of the forward-haul rate. In so doing, the Commission purported to act under its § 4 power. In a § 5 proceeding, of course, the Commission is without authority to order refunds. See 15 U.S.G. § 717d(a) (section 5 allows FERC to “determine the just and reasonable rate ... to be thereafter observed and in force”) (emphasis added); see also Arkansas Louisiana Gas Co. v. Hall, 453 U.S. 571, 578, 101 S.Ct. 2925, 2930-31, 69 L.Ed.2d 856 (1981); Tennessee Gas Pipeline Co. v. FERC, 860 F.2d 446, 454 n. 10 (D.C.Cir.1988); Sea Robin Co. v. FERC, 795 F.2d 182, 189 n. 7 (D.C.Cir.1986).
We are aware that the Commission’s order of refunds under § 4 is discretionary. See Beko Petroleum v. FERC, 589 F.2d 680, 686 (D.C.Cir.19,78). “In granting the Commission that discretion, Congress did not specify how the refunds should be computed; this, too, was left to the discretion of the agency.” Id. Pursuant to this discretion, however, the Commission’s longstanding policy has been to grant full refunds of rates found unjustified at the conclusion of § 4 proceedings, in the absence of equitable considerations to the contrary. See Towns of Concord, Norwood & Wellesley v. FERC, 955 F.2d 67, 76 (D.C.Cir.1992) (citing Illinois Power Co., 52 F.E.R.C. ¶ 61,162, ¶ 61,625 (1990)); see also Estate of French v. FERC, 603 F.2d 1158, 1163 (5th Cir.1979) (enumerating countervailing equitable considerations such as passage of time, size of producer, magnitude of amounts owed, public interest). Presuming no countervailing equitable concerns in this case, we would expect that the Commission should order refunds of rates collected in excess of the preexisting filed rate, rather than in excess of the 50% back-haul rate it imposed. Because, however, we hold that the Commission’s action is presently supportable under neither § 4 nor § 5, we remand to the Commission for reexamination of the refund issue in light of its reconsideration of the forward-haul and backhaul rates themselves. At this juncture we note only that if the Commission proceeds to impose a new rate under § 5, after rejecting a § 4 rate proposal as unjust and unreasonable, we see no statutory impediment to the Commission’s exercise of § 4 refund authority to prevent the pipeline from benefitting from an unjust or unreasonable rate in the interim before remedial action is taken.
III. Conclusion
For the foregoing reasons, we remand to the Commission so that it may more adequately explain its conclusion approving the proposed forward-haul rate increase and revisit its actions with respect to the backhaul rate.
It is so ordered.
. The agency proceedings refer to Western as the Kansas Power & Light Company, its official name until May 1992. For the sake of clarity, we refer to the company throughout as ‘‘Western.”
. The contract between Western and Panhandle has expired since the issuance of the Commission’s decision.
.As we noted in ANR Pipeline Co. v. FERC, 771 F.2d 507 (D.C.Cir.1985), a "[bjackhaul is accomplished not by actually transporting gas back west (which would require a separate pipeline) but rather by performing a swap. When ANR wishes to ‘move’ gas from east to west, gas is removed from the eastward flowing stream at the western destination point. ANR then ‘pays back’ the pipeline by restoring the appropriate quantity *1571of gas to the gas stream at the eastern origination point.” Id. at 511-12.
. The contract defines the term "Mcf” as one thousand cubic feet of gas. Joint Appendix ("J.A.”) at 66.
. According to Western, rates under Rate Schedule T-53 increased from $907,220 to $2,935,651, at 224% leap. See J.A. at 43.
. This lack of price-flexibility is not surprising. As we noted in Columbia Gas Transmission Corp. v. FERC, 848 F.2d 250 (D.C.Cir.1988), "FERC typically has refused to authorize selective discounting under individual section 7(c) certificates.” Id. at 254.
. Like the airline passenger who purchases a nonrefundable airline ticket months in advance— for an off-peak, nonblackout period, no less— Western expects that the significant limitations on its service should be reflected in the price it pays. Its contention at this juncture, that the off-peak limitation should translate into less costly service, has inherent appeal.
. As we have mentioned previously, although 15 U.S.C. § 717r(b) requires that "[t]he finding of the Commission as to the facts, if supported by substantial evidence, shall be conclusive,” this is merely a recitation of the "arbitrary and capricious” standard for factual findings. See Tennessee Gas Pipeline Co. v. FERC, 860 F.2d 446, 452 n. 7 (D.C.Cir.1988); Maryland People's Counsel v. FERC, 761 F.2d 768, 774 (D.C.Cir.1985).
.In this respect, the Commission seems to have verged precipitously close to begging the question. The issue reserved for the Commission’s determination was whether, given the distinctions between these two services, it was appropriate to price them equally. It is a far cry from responding, surely, to conclude that the distinc*1575tions themselves should not exist. Moreover, it strains Western's argument somewhat to depict it as an attempt to procure a less-constrained T-53 service.
. We carved exceptions, inapplicable here, for situations in which Congress specified that courts should play an independent role and situations in which the agency itself is an interested party to the agreement. The latter exception seems to contemplate only the extreme case in which FERC itself is a party to the original contract. See National Fuel, 811 F.2d at 1571.
. Specifically, the statute provides an exception to the rule prohibiting retroactive rate changes "in order to accommodate the realities of administrative delay.” East Tennessee, 863 F.2d at 942. If a FERC proceeding extends over five months, the statute permits the pipeline to collect the proposed rates on a temporary basis. If the Commission ultimately concludes that the proposed rates are not just and reasonable, it may require the pipeline to pay refunds. We recognized a further exception to the general rule against refunds in Cities of Batavia v. FERC, 672 F.2d 64 (D.C.Cir.1982). This exception subjects a pipeline to refunds if the interaction between proposed and existing rates will create results that are unjust or unreasonable under existing FERC policy as it applies to the pipeline at the time of filing. See East Tennessee, 863 F.2d at 943.
. In which, it should be recalled, the forward-haul rate was higher by a factor of twenty.