CALC-063

Every other institutional estimate restated by the CRF shortcut

Feeds CLM-0370, CLM-0465, CLM-0372 (low confidence), CLM-0374 (low confidence) and CLM-0375 (low confidence). SHARED - five claims, one re-runnable section. This is the largest reduction in this half of the catalogue. Answers QST-004. Basis: GBP2024/MWh, 20-year asset life, 3 and 10 per cent real, except where a row says otherwise.

Script. restate-lcoe.py SECTION 5, section_restate_all() (line 400) with restate() (line 356). One function, five studies, ten-plus output rows.

Why one record and not five. The natural re-runnable unit is the function, not the study. Every row below is the same three operations in the same order - inflate in native currency, convert at 2024 FX, rescale by CRF ratio - differing only in the constants fed in. Drafting five records would be drafting one method five times and would hide the fact that a single defect in restate() moves all five together. What must not be lost by sharing is the per-study input status, so it is given in full below rather than summarised.

Inputs, per study. The `Stated or assumed` column is the load-bearing one.

Working, transcribed.

-- Frazer-Nash BEIS 2021, FOAK p50 (CLM-0370) --
GBP2018 66.30 x 1.2658 (UK CPI to 2024)      = GBP2024 83.92 /MWh
             x 0.3347  (CRF ratio, 20%/30yr -> 3%/20yr)  =  28.1
             x 0.5848  (CRF ratio, 20%/30yr -> 10%/20yr) =  49.1
NOAK 5th, GBP49.70 native   ->  21.1 (3%)  /  36.8 (10%)

-- Frazer-Nash ESA 2022, FOAK p10 (CLM-0465) --
EUR106.00 x 1.0793 (EA HICP to 2024)         = EUR2024 114.41 /MWh
          x 0.8466 (GBP per EUR, 2024)       = GBP2024  96.86
          x 0.3347 (CRF ratio)               =          32.4
FOAK p10-p90   32.4 - 63.0  (3%)   /   56.6 - 110.1 (10%)
NOAK 10th p50-p90  27.1 - 47.6 (3%)  /  47.3 -  83.1 (10%)

-- Roland Berger / OHB 2022 (CLM-0372) -- CURRENCY ONLY, NOT RESCALABLE --
EUR69.00 x 1.0793 = EUR2024 74.47 x 0.8466 = GBP2024 63.05 /MWh FOAK
                                              GBP2024 44.8  /MWh NOAK
NO discount-rate or asset-life rescaling is possible. The source states
neither parameter. These two figures are NOT on the common basis.

-- NASA OTPS "competitive" sensitivity (CLM-0374) -- CRF shortcut --
RD1  33.1 (3%)  /  57.9 (10%)      RD2  88.3 (3%)  /  154.4 (10%)

-- Caltech CSSPS, three tiers (CLM-0375) --
current      2,647.6 (3%)  /  4,626.6 (10%)
middle          48.4 (3%)  /     84.6 (10%)
asymptotic      22.7 (3%)  /     39.6 (10%)

Result: eleven restated figures spanning GBP21.1 to GBP2,647.6/MWh at 3 per cent real, all GBP2024 per MWh at 20-year life, except the two Roland Berger figures, which are on a currency-converted basis only.

Conclusion, and the strongest single item in this record is a refusal. CLM-0372 (low confidence) is the clearest instance in the whole restatement of "cannot be fully restated", and the claim was written to say so rather than to invent the two missing parameters. A missing parameter recorded as missing is a finding. The alternative - assuming 3 per cent and 30 years to make the row fit the table - would have produced a number indistinguishable in the output from the four that are real.

Where this fails, per study, because the failures are not shared even though the method is. - Frazer-Nash BEIS and ESA. CRF shortcut, not a rebuild; carries CALC-061's plus or minus 25-30 per cent. No bottom-up CASSIOPeiA model exists in the registered evidence. The probabilistic launch-cost distribution behind the native BEIS figure could not be re-run at the new basis. - Frazer-Nash ESA additionally. A 1.44 GW system against NASA's approximately 2 GW; not size-comparable without an adjustment that cannot be made here (CALC-068). - Roland Berger. A VALCOE, not an LCOE. That is a scope difference on top of the two missing parameters, and its direction is stated: netting dispatchability value out typically makes a VALCOE lower than the equivalent plain LCOE, not higher. Cannot be compared like-for-like with any other row. Its only source is a T5 marketing or insights web page. - NASA competitive sensitivity. NASA's own stated best case, stacking five or six simultaneous, individually under-specified improvements (the report's own lists differ; CLM-0124). Not a forecast. Must never be quoted without that context, per CLM-0374 (low confidence)'s own limits. - Caltech. The asset-life substitution is the material one: the "N2" multiplier is used as a life term although the source does not say it is one, and every figure in the row moves with it. The middle and asymptotic tiers additionally rest on launch prices of approximately 128 and 55 USD/kg projected from a 2020 planning document and never observed at any date. 76-181 MW average power, one to two orders of magnitude below every other study.

A LOCATOR REPAIR ALREADY MADE IN CLM-0465, CARRIED HERE SO IT IS NOT LOST. The 30-year satellite life for the ESA study was originally cited to p.55 of SRC-078. p.55 covers an unrelated funding-split table; the figure is on p.58. Found under VER-201 and repaired 2026-09-03. Every restated GBP figure reproduced exactly from a fresh run of the script, so this was a citation defect and not an arithmetic one - the same pattern as the four locator repairs recorded at CALC-056.

One transcription note on CLM-0465, offered as a reading. The claim carries p10 and p90 separately and states why: "no single central (p50) FOAK figure was stated by the source, so p10 and p90 are carried separately rather than an invented midpoint." That is the correct handling and it is worth naming, because the alternative - averaging to a midpoint the source never published - is how a fabricated central estimate enters a register.

Cite this