Mike, your example does not take into account engine loading which varies constantly in the real world. As one example, 2k rpms up a steep hill pulling a trailer is not the same as 2k rpms on flat ground. The throttle position alone tells you one scenario is consuming way more fuel for the same engine rpm compared to the other one.
In order to hit 1% as a SYSTEM, the sub-component tolerances and variations stack up must not total more than that 1%. I don’t believe the odometer alone is even better than 1% and that leaves a boat-load of other factors un-accounted for in the error budget as I mentioned previously.