I think no one here really knows ... the truth seems to remain a mystery ... but they sound like good guesses tho ...
the only accurate way to get an idea of the amount of energy spent moving the mass of the drive train + the other factors such as heat, friction etc (as suggested) would be to do several tests comparing the engine and engine drive train combo <--- which no one is likely to try ... and that would only be reasonable if the next car in question had the same setup as the original test car/engine & drive train combo.
the hypothesis that the drive train under full load will rob a fixed amount of power + plus a proportional amount depending on load increase/ power + other factors such as fixed heat dissipation etc sounds the most accurate but it is probably much more complicated than that.
anyway the thread was a good read got me thinking ...