Cant be bothered to try and figure out the maths... its too hot in this office!
If we're trying to get figures as close to 'at the crank' as possbile, then short of getting the engine out, you want to try and remove any of the inacuracies of a dyno or RR as possible. An RR will give a certain amount of wheel slip, this would be unmeasurable in real terms, and as such only a 'fudge factor' could be applied to give a 'crank figure'. A hub dyno removes wheel slip and would therefore be slightly more accurate, although you would still have to apply a guestimate on the transmission losses, and even then each cars' transmission losses would differ.
Am I making sense or talking out of my arse!