Gareth
They're subject to the same limitations as any other compiler, in that going allowances are - at best - an educated guess, and the 12f h/c, Cadran and Opera are no help in assessing that figure.
They may not have given a per furlong figure for the 7 & 8f races, but I'd doubt they've "given up altogether" as it's just a simple division of the time by the distance, and all 3 races can be done in seconds by any 6yo with a calculator, so I'd seriously doubt they've used any fewer races than other compilers.
They also 'may' have assumed the Arc was truly run, so (probably) have most other clockers - but the sectionals show that wasn't necessarily true, so maybe they all guessed wrongly.
RP standard times may not be the best in the world, but they're certainly not as black as you've painted them, and while I believe that no ratings can tie a horse (or a going allowance) down to the nearest pound, they'll service adequately - in the circumstances I described earlier.
Even ignoring all the foregoing, the visual evidence of Orfevre quickening strongly suggests the ground was no worse than soft, and I've yet to see anyone suggest a horse showing similar acceleration on anything worse.