Opinion for the Court filed by Circuit Judge STEPHEN F. WILLIAMS.
STEPHEN F. WILLIAMS, Circuit Judge:
When a utility files a proposed rate increase with the Federal Energy Regulatory Commission, it supports the filing with current estimates of its future costs and revenues. The Commission normally bases its decision on the estimates’ reasonableness when made, disregarding events occurring between the filing and its own decision, and shunning the model of Penelope, who nightly unwove all that she had accomplished during the day. Thus it appears to regard the delay and administrative costs of incessant updating as too great to justify whatever increased accuracy might result. Moreover, the yield in accuracy might be small, as errors upward in one estimate may be offset by errors downward in another. See Carolina Power & Light Co. v. FERC, 860 F.2d 1097, 1102 (D.C.Cir.1988). Where challengers of the proposed rate show that use of the estimates would be “unreasonable”, however, even though the estimates were reasonable when made, the Commission will order a “spot adjustment” in the light of more current data. See, e.g., Villages of Chatham v. FERC, 662 F.2d 23, 29 (D.C.Cir.1981).
The decision under review tests some of the limits of this exception. The utility, Southwestern Public Service Company, here claims that a spot adjustment is never appropriate for changes in tax rate occurring after the close of the period used for estimating costs. It also claims that even if it were, the Commission erred in making an adjustment without considering other facts that made the proposed rate reasonable in the face of changed tax rates. We find the first argument unsupported; we remand on the second because the Commission has failed to explain adequately its refusal to consider the impact of Southwestern’s alleged offset.
On May 1, 1985 Southwestern filed a proposal to increase the rates at which it sells electricity wholesale to several municipalities and electric cooperatives. See § 205 of the Federal Power Act, 16 U.S.C. § 824d (1982). It submitted not only historical data for a past 12-month period (“period I”) but also an estimate of the costs and revenues for a future 12-month period (“period II”), and the Commission used the latter for determining Southwestern’s rates. See 18 CFR § 35.13(d)(4) (1990). For period II, Southwestern selected the 12 months that began on July 1, 1985 and ended on June 30, 1986. It proposed that the rates become effective January 21, 1986.1
While the filing was before an administrative law judge, Congress changed the applicable corporate income tax rates. The top rate had been 46% at the time of filing, and Southwestern had assumed that that rate would continue for all of period II. And so it did. But on October 22, 1986— shortly after the end of period II and shortly before the AU issued his opinion on November 6, 1986 — Congress enacted the Tax Reform Act of 1986, Pub.L. No. 99-514, § 601(b), 100 Stat. 2085, 2249 (1986), reducing the rate to 34% beginning July 1, 1987. Thus, although Southwestern’s period II estimate of its tax costs remained accurate for period II itself, that figure greatly exceeded the costs likely for the last two and a half years of the period for which the rates were effective — from July 1, 1987 to its close on December 18, 1989.2
[201]*201Accordingly, in the course of an appeal to the Commission, Southwestern’s customers3 and the FERC staff asked FERC to re-open the record and to make a spot adjustment of the period II estimates in order to reflect the reduced tax costs. FERC did not act on the motions to re-open the record, but it agreed in substance, ordering Southwestern to submit new analyses of costs for three separate periods: (1) from January 21, 1986, when Southwestern’s new rates became effective, to August 31, 1986, at the existing 46% tax rate; (2) from September 1, 1986 until August , 31, 1987, Southwestern’s next taxable year, at an appropriate “blended” tax rate that would take the July 1, 1987 tax reduction into account; and (3) for Southwestern’s taxable years after September 1, 1987, at the new 34% tax rate. Southwestern Public Service, 49 FERC ¶ 61,354 at 62,283-84 (1989) (“Order ”). On Southwestern’s petition for rehearing, FERC stuck to its view that the change in tax rate rendered the period II estimates unreasonable, and also rejected Southwestern’s argument that the spot adjustment should in any event be offset by increases in the cost of power that Southwestern purchased from Public Service Company of New Mexico (“PNM”). Southwestern Public Service, 53 FERC ¶ 61,084 (1990) (" Order on Rehearing ”). We address first Southwestern’s attack on the propriety of a spot adjustment based on substantial tax rate changes occurring after the close of period II, and then its view that because of the increased PNM costs, its estimates were reasonable despite the tax change.
The Commission has stated that spot adjustments are appropriate when “it can be demonstrated that the estimates were either unreasonable when made or, if reasonable when made, subsequent events indicate that to use them as a basis for future projections would yield unreasonable results.” Southern California Edison Co., 8 FERC ¶ 61,099 at 61,375 (1979). Later events render an item unreasonable only if they show it to be “substantially in error”. Id. (emphasis in original). The formulation of the first issue — whether the estimate was reasonable when made — appears to be no more than what the Commission requires of a utility regardless of intervening changes. As there is no dispute over the estimates’ initial reasonableness, however, we need not worry about whether this seeming duplication has any significance in the real world.
At first sight, a substantial reduction in tax rates appears to qualify as a change that would presumptively render the previously filed rates unreasonable. Southwestern, however, claims that FERC has a general policy against treating tax rate changes as the kind of event that warrants a spot adjustment, and points to Public Service Co. of New Mexico, 10 FERC ¶ 61,053 (1980), as clear evidence of this policy. There FERC refused to order a spot adjustment to reflect a .post-period reduction in the federal corporate income tax rate from 48% to 46%, declaring that, under its policies, “there is no opportunity to go beyond period II to adjust for known and measurable changes except in the area of rate of return.” Id. at 61,123.4
The trouble with Southwestern’s argument is that FERC’s language in Public Service Co. of New Mexico was drastically and clearly in error. FERC had only recently restated its policy permitting spot adjustments to reflect changes in any area of costs or revenues, so long as the changes were unforeseeable and “substantially in error” and would yield unreason[202]*202able rates. See Southern California Edison, 8 FERC at 61,375 (emphasis in original). It seems plain that the Commission regarded the 2% tax rate change in Public Service Co. of New Mexico as insubstantial.
There was a germ of truth, however, behind Public Service Co. of New Mexico’s mistaken dictum. The Commission does draw a distinction between changes in rate of return and all other changes, allowing a spot adjustment for the former even where the change is slight. And in Public Service Co. of New Mexico it explicitly made the case for such costs’ singularity, quoting at length from an earlier decision explaining that view, Union Electric Co., 47 FPC 144 (1972). While other cost changes tended to be offset at least partially by “countervailing changes in technology, efficiency, and market growth”, capital costs tended to “rise and fall rather sporadically ... without equally visible and convincing evidence of offsetting cost trends.” 47 FPC at 156. While in Union Electric the Commission did not characterize the rate of return change as large or small, that very omission suggested its indifference to scale for those changes. At the same time, the Commission emphasized both the singularity of rate of return and need for substantiality for other costs by noting that it would “continue to resist piecemeal, post-test year changes in other cost factors [other than financing costs]”. Id. (emphasis added). Thus Public Service Co. of New Mexico appears merely to endorse FERC’s established doctrine that rate of return changes may give rise to spot adjustments regardless of size, rather than some novel rule that tax rate changes cannot trigger an adjustment even if substantial.
In sum, Public Service Co. of New Mexico and Southern California Edison make sense as reflecting two independent exceptions to the general policy against spot adjustments of estimates that were reasonable when made. Spot adjustments may be used to reflect (1) changes in the rate of return regardless of scale, and (2) substantial deviations between estimates and later data that would result in unreasonable rates. The two exceptions do not conflict; they exist side by side. Thus in Delmarva Power & Light Co., 22 FERC ¶ 63,052 at 65,185 (1983), rev’d on other grounds, 770 F.2d 1131 (D.C.Cir.1985), the Commission rejected the utility’s request to include post-test period costs of a new plant, first because “with the exception of data relating to rate of return, post-test year changes are generally unacceptable”, and second because the estimate was not substantially in error. In other words, the changes fitted neither exception.
Despite the apparent pattern, Southwestern points to Williston Basin Interstate Pipeline Co., 52 FERC ¶ 61,170 (1990), as evidence that the Commission either in reality adheres to the mistaken dictum in Public Service Co. of New Mexico or, at best, has no policy other than total disarray. In that case FERC reversed the AU’s order to a natural gas pipeline to make spot adjustments for changes in costs resulting from the Tax Reform Act of 1986, and, as in Public Service Co. of New Mexico, used very broad language as a purported explanation: “Commission policy is generally not to adjust individual items of the cost of service to reflect changes that occur after the close of the test period,” id. at 61,645. It went on to say that it had previously ordered changes on account of the 1986 Act where the effective date of the tax reduction came “within, or soon after the close of,” the test period. Id. As Williston’s test period here ended 15 months before the 1986 Act’s effective date, those cases did not apply.5
As the interval in this case is 12 months rather than 15, one could reconcile the two cases on that ground, though perhaps only if FERC itself explicitly drew the line. But [203]*203Williston Basin also noted that the “rates at issue here were in effect only until February 29, 1988 — 8 months after the effective date of the tax reduction.” Williston Basin, 52 FERC at 61,645 n. 26 (emphasis added). FERC’s emphasis on the shortness of the period for which the arguably obsolete rates were locked in suggests that it regarded the problem as involving only a “piecemeal” change in costs.
The Commission also draws some support from Kansas Gas & Electric Co., 49 FERC ¶ 61,295 (1989), decided two weeks before its original order here. There too it followed Southern California Edison, ordering a spot adjustment on account of the 1986 Act, even though it was enacted after the end of the test period, and arguing first that “the effect of the change in the corporate tax rate is known and measurable, certain in amount, and not subject to estimation”; and second that the effect of the change in tax rates is “not insignificant”. 49 FERC at 62,123.
Another stray remark of the Commission requires explanation. In Delmarva, 22 FERC ¶ 63,052 at 65,185 (1983), the Commission had refused a spot adjustment for a post -period change, distinguishing Public Service Co. of Indiana, 57 FPC 1173 (1977), aff'd in part, 575 F.2d 1204 (7th Cir.1978), on the ground that it had made an adjustment to reflect experience during the test period. This may have seemed a suggestion that post-period changes could never give rise to a spot adjustment. But Delmarva also explained that, “[e]ven if PSC of Indiana were applicable,” it would do Delmarva no good as the post-period change was insubstantial. 22 FERC at 65,185. In fact the Commission has on occasion made spot adjustments for events occurring entirely after the test period. See, e.g., Distrigas of Massachusetts v. FERC, 737 F.2d 1208, 1220 (1st Cir.1984) (adjusting estimates in light of actual figures for a year starting after the nine-month “adjustment period” (a period used in gas cases and itself starting after the test period)); NEPCO Municipal Rate Committee v. FERC, 668 F.2d 1327, 1340 (D.C.Cir.1981) (adjusting estimate of test period cost in light of estimates of costs for a year after the test period); Northwest Central Pipeline Corp., 38 FERC ¶ 61,171 at 61,556 (1987) (gas case where utility used actual experience for test period). Given this precedent, the absence of any apparent reason why adjustment for a post-period change would cause any more difficulty than adjustment for an in-period change, and the absence of any FERC decision clearly supporting the distinction, we find no basis for this limitation. See Villages of Chatham, 662 F.2d at 35 (FERC has discretion to consider post-period II figures); but cf. Virginia Electric, 10 FERC ¶ 61,083 at 61,180 (1980) (FERC grants a spot adjustment for the 48% to 46% tax change at around the same time as Public Service Co. of New Mexico, the only difference being that the tax reduction occurred during the test period in Virginia Electric, not after the test period as in New Mexico); Delmarva, 770 F.2d at 1139 (remanding order forbidding selective annualization as inadequately explained, in dicta suggesting significant difference between test period and post-test period events).
Finally, Southwestern argues that FERC’s order violates Carolina Power & Light Co. v. FERC, 860 F.2d 1097 (D.C.Cir.1988). In that case the utility’s test period was calendar year 1987 and the date of its rate increase was September 1, 1987. The utility had assumed a blended federal rate of 39.95% (46% and 34%, weighted) for the test year. FERC ordered the utility to adjust its estimate to reflect a tax liability of 34% for the remaining four months of 1987 (the utility having agreed with FERC to file new rates effective January 1, 1988 reflecting the new 34% rate). This court held- FERC’s order to be arbitrary and capricious, because the 34% figure would be clearly below the blended rate of 39.95% that CP & L was required to pay. Although this left CP & L with a windfall for the first eight months of 1987, the court rejected FERC’s claim that that justified the order, as it appeared to be simply an indirect refund for excess income collected in a period covered by rates previously approved and for which no refund was [204]*204lawful. Id. at 1101-02; cf. id. at 1103-04. FERC had done indirectly what it could not do directly. See id. at 1101.
Because FERC had defended its action under the spot adjustment doctrine, the court then went on to discuss that — although no more was necessary beyond the point that the spot adjustment doctrine cannot shelter a decision to order otherwise illegal refunds for prior periods. In doing so, the court read the broad language in Public Service Co. of New Mexico and one of its clones (Central Illinois Light Co.) to reflect some sort of general Commission rule against spot adjustments for changes in income tax rates, id. at 1102, and even went on to suggest that perhaps such a rule was legally necessary, id. at 1102-03. In seeming contradiction of that dictum, it went on to suggest that FERC might not have regarded the change (39.95% to 34%) as “substantial enough” for a spot adjustment, id. at 1102, citing an AU’s decision refusing a spot adjustment in Kansas Gas & Electric Co., 39 FERC ¶ 63,013 at 65,070 (1987)—an AU decision later reversed by the Commission in its 1989 decision discussed above. While these dicta are far from clear to us, their status as dicta is clear. We find them no barrier to the Commission’s decision here to apply its regular rule on spot adjustments to changes in income tax rates.
Although there is some disarray among FERC and circuit court precedents, the path FERC chose was not arbitrary.
Southwestern argues that even if a spot adjustment is normally permissible for a tax change such as that enacted in 1986, it should not be made if the divergence between estimates and reality on tax costs is offset by an equally great divergence on another item. The mistakes balance out and the rates are still reasonable. Here, it says, the windfall from the unexpected tax reduction was offset by higher costs resulting from increases in the price of fuel under its contract with PNM — increases that the contract specifically required but that (for reasons discussed below) were not encompassed in the cost estimates. Therefore, Southwestern argues, the rates set prior to these events continue to reflect its total costs, and a spot adjustment is improper.
As we have seen, FERC has resolved the trade-off between the interests of speed and simplicity, on the one hand, and up-to-the-minute accuracy, on the other, with provision for spot adjustments only for post-filing changes (other than rate of return changes) so substantial that they would render the rate unreasonable. But the Commission’s resolution of this threshold question gives rise to another: in cases where the threshold is crossed, when (if ever) should the Commission consider offsetting changes alleged to have salvaged the rate’s reasonableness? Broadly speaking, there are three possibilities. At one extreme, the Commission might consider all offsetting costs, on the ground that failure to do so could lead to “unreasonable rates”; but this would risk undoing the speed and simplicity sought by the basic rule. At the other extreme, it could refuse to consider any offsetting changes — or at any rate any except for ones that themselves qualified to trigger a spot adjustment. This would preserve speed and simplicity at a high cost in accuracy (at least the cost will be high if a non-trivial share of substantial changes are in fact offset). The PNM changes would not qualify under this view, for though evidently substantial, they were not unforeseeable at filing but were embedded explicitly in a signed contract. Finally, FERC could consider some but not all intervening changes that fall short of what is necessary to justify a spot adjustment. Southwestern asserts without contradiction that the PNM changes are very like the tax changes except as to foreseeability — they are “known and measurable, certain in amount, and not subject to estimation”, and “not insignificant”, see Order, 49 FERC at 62,283 (speaking of the tax rate changes). It says FERC should at least consider changes such as these.
Southwestern reads the Commission’s prior decisions as supporting consideration of offsets generally, or at least of ones as substantial and discrete as both the PNM [205]*205and tax rate changes. FERC, by contrast, reads the same cases as embracing only offsets in costs “related” to the change prompting the adjustment. FERC Br. at 31-33. In fact, neither of the FERC orders under review mentions this “relatedness” theory, so it is most doubtful whether we could affirm on that basis even if the precedents supported FERC’s theory. See SEC v. Chenery Corp., 332 U.S. 194, 196, 67 S.Ct. 1575, 1577, 91 L.Ed. 1995 (1947). In any event, we can find no such rule in the cases the Commission cites.
FERC’s prime cite is Public Service Co. of Indiana, 57 FPC 1173 (1977), aff'd in part, 575 F.2d 1204 (7th Cir.1978), which in fact says nothing of relatedness. It observes, in language quoted by the Commission, that insubstantial changes are to be disregarded because of the likelihood of offsetting changes. FERC Br. at 31-32 (quoting 57 FPC at 1183). On its face this says nothing about how to handle possible offsets once an unforeseeable and substantial change is identified. In fact, no one appears to have asserted an offsetting change, and the Commission did not even offer a dictum on the subject.
The Commission nonetheless cites Public Service Co. of Indiana as a case where offsets were taken into account (only because related), identifying the matching items as “purchaser power costs and short-term resale revenue”. A glance at the decision makes clear first that the Commission allowed the spot adjustment (so it cannot be a case of admitting “related” offsetting changes), and second that the Commission did so for the simple reason that it had only one change before it — the utility’s unexpected receipt of about $15 million in resale proceeds. 57 FPC at 1183-84. These receipts balanced an expected $15 million in purchased power costs, and the Commission effected the adjustment by eliminating the $15 million in costs, but its decision to do so was a formality of no consequence, certainly with no implication that there were two changes involved. Commission counsel have got this case backwards.
The Commission also tries to classify Southern California Edison Co., 8 FERC ¶ 61,099 (1979), as a case where offsets were allowed only because the items were related — “load factor and sales revenue”. FERC Br. at 32. Here the Commission denied the proposed spot adjustment, but not because of an offset of related costs. It denied it simply on the ground that the post-filing “actual” experience (load factor) for the test period was unrepresentative, while the test-period estimate was correctly “based on recent trends”. 8 FERC at 61,376. The relation of load factor and sales revenue came into it only in the sense that one determined the other: under the applicable rate design, “an underestimation of the load factor results in excess revenues.” Id. (emphasis added).
Finally, the Commission invokes Indiana Municipal Elec. Ass’n v. FERC, 629 F.2d 480 (7th Cir.1980), which, unlike its other cites, actually uses the word “unrelated”. But the context makes clear that the court was concerned with a much more concrete problem. In upholding the Commission’s refusal to make a spot adjustment where certain unexpected revenues were roughly matched by unexpected costs, the court observed:
These [revenues and expenses] were both wholesale revenue and expense figures, so that the Commission did not offset unrelated accounts or use retail and wholesale figures as the Association has contended.
Id. at 484 (emphasis added). See also id. at 485-86 (similar). Thus the court was evidently responding to the complaining wholesale customers’ argument that they had been wrongly denied a spot adjustment because of costs unrelated to wholesale service. We assume the Commission would reject retail service costs as an offset to wholesale revenues, but that rejection says nothing of the concept its attorneys now discern.
It is true that in the order underlying the Seventh Circuit’s decision the Commission had suggested the possibility of a limitation that parallels counsel’s current theory. The utility had argued that the impact of [206]*206unexpectedly lower costs would “be negligible if the offsetting impact of certain (assertedly ‘synchronized’) adjustments is also taken into consideration.” Public Service Co. of Indiana, 7 FERC ¶ 61,319 at 61,702, reh. denied, 8 FERC ¶ 61,224 (1979). The Commission accepted the utility’s claim that there was a valid offset and observed in a footnote:
We ... note ... that evidence of overall actual costs or revenues is not necessarily synonomous [sic] with evidence of “synchronized adjustments.” “Synchronization” means adjusting only the particular line items — such as demand allocation or fuel cost — which are in fact directly or intimately related to the challenged primary estimate or estimates of which adjustment is sought.
Id. at 61,712 n. 18. The utility explained that alleged “synchroniz[ation]” (a term likely used by it and the Commission for what the latter now calls relatedness) came through a single dominant cost: “The two figures, lower revenues and lower expenses, are directly related through the cost of fuel____” Id. at n. 16 (quoting utility brief).
But FERC simply ruled against a spot adjustment, relying on the utility’s schedule of its “overall actual jurisdictional expenses and revenues”, id. at 61,702 (emphasis added), and foregoing any evaluation of the assertion of synchronization. Presumably that reflected its judgment that no such inquiry was necessary; to the extent that the Seventh Circuit’s reference to relatedness bears on anything other than the retail/wholesale distinction, it cannot manifest reliance on an assumption that the Commission performs such an inquiry.
Thus the . Commission’s decisions fail to establish any policy that a substantial error can be offset (and a spot adjustment negated) only by changes in “related” costs. Moreover, we have at least twice expressed our understanding that under Commission policy the challengers of a utility’s estimates, once the utility has shown them to be reasonable when made, must carry the burden of showing not only a substantial error but also that the resulting disparity “would yield ‘unreasonable results.’ ” Boroughs of Ellwood City v. FERC, 731 F.2d 959, 966 (D.C.Cir.1984); see also Indiana & Michigan Municipal Distributors Ass’n v. FERC, 659 F.2d 1193, 1198 n. 14 (D.C.Cir.1981) (“challengers must go beyond establishing a substantial disparity between the test-year and historical figures. They must proceed to establish that the use of the estimate would yield, overall, an unreasonable result by showing, for example, that there were no offsetting considerations or that the causes underlying the increase in short-term sales could be expected to spill over into the remaining months covered by the rate increase.”) Unless the Commission’s alternative theory on the PNM costs (addressed below) carries the day, the complaining customers have not yet met that burden.
FERC’s alternative theory, advanced in the Order on Rehearing, 53 FERC at 61,243, is that any costs resulting from the PNM contract were settled in other proceedings, so that Southwestern’s effort to use them now represents an invalid “collateral attack” on the earlier proceedings. This may be so, but the record before us is too unclear to show it.
Southwestern’s contracts with its customers allowed for automatic adjustment of its rates to reflect its changing fuel costs. See Southwestern Public Service Co., 37 FERC ¶ 63,012 at 65,120 (1986). But as FERC regulations limited the use of fuel adjustment clauses to purchase contracts of one year or less, see 18 CFR § 35.14 (1990), this would not cover the costs under its five-year contract with PNM. Accordingly, in its 1985 filing, Southwestern asked for a waiver of the one-year maximum. In a preliminary ruling, the Commission rejected the request. Southwestern Public Service Co., 32 FERC ¶ 61,249 at 61,584 (1985). But in later proceedings on some customers’ claims that Southwestern manipulated the allocation of fuel costs between its on-system and its off-system customers, an AU saw the fuel adjustment clause as a solution: Southwestern would be allowed to use the clause for its PNM costs, but the [207]*207proceeds of any off-system sales would also be flowed into that calculation, so that any net benefit from the sales would go to on-system customers. Southwestern Public Service Co., 37 FERC at 65,122.
FERC vacated the ALJ’s remedy on the grounds that the record was inadequate for evaluating the issue, and that it would be addressed more appropriately in a hearing on a separate complaint filed by one of Southwestern's customers. Order, 49 FERC at 62,280; see also Golden Spread Electric Cooperative, Inc. v. Southwestern Public Service Co., 49 FERC ¶ 61,364 at 62,306-07 (1989) (ordering hearing on Golden Spread’s complaint). Southwestern did not appeal. Golden Spread was settled, see Order on Rehearing, 53 FERC at 61,243, and the Commission views Southwestern’s effort to use the PNM costs here as a collateral attack on that settlement.
The relation of the complaint proceeding to the present issue is obscure. The Commission ordered the Golden Spread complaint resolved by use of a test year of calendar 1990, and selected as the refund date the date of issuance of its decision on Southwestern’s 1985 filing (December 19, 1989), see Golden Spread, 49 FERC at 62,306-07, so the case appears to cover a different time period. As the Commission notes, however, the settlement included a dismissal by the complaining customers of their complaints about Southwestern’s fuel pricing for the years 1985-90, see FERC Br. at 34, Southwestern Br. at 29 n. 40, suggesting both that the period at stake here was involved and that perhaps Southwestern has already bargained away its possible claims to higher charges for PNM power in 1985-89. Using them to negate a spot adjustment would thus be a double counting. At present, however, the record is inadequate to draw this conclusion.
Accordingly, we must remand the case to the Commission for it to address the question of whether the customers have shown that, despite the PNM costs, the 1986 tax rate changes rendered Southwestern’s rates unreasonable. A clear analysis of the Golden Spread settlement may dispose of the problem. If not, the Commission may wish to impose a “relatedness” criterion on the offsets that may negate a spot adjustment; the issues implicit in such a judgment are not before us. If the Commission does establish such a criterion, it will have to address Southwestern’s contention that, since no such doctrine existed during the pendency of its filing, the Commission’s prolonged inaction lulled it into foregoing the opportunity to recoup those costs by filing new rates under § 205, inaction that proved fatal under the decision that finally emerged from the Commission. Cf. Chevron Oil Co. v. Huson, 404 U.S. 97, 92 S.Ct. 349, 30 L.Ed.2d 296 (1971); American Gas Ass’n v. FERC, 888 F.2d 136, 150 (D.C.Cir.1989). If FERC does adopt the relatedness criterion and concludes that it may retroactively apply the criterion to Southwestern, it should take note of our recent suggestion that James B. Beam Distilling Co. v. Georgia, — U.S. -, 111 S.Ct. 2439, 115 L.Ed.2d 481 (1991), may forbid agencies to apply rules with selective retrospectivity. See District Lodge 64 v. NLRB, 949 F.2d 441, 446 (D.C.Cir.1991). Accordingly, the case is
Remanded.