Fred, et al,

 

There is at least one more factor to consider - Timing (assuming these engines are all running in dry air).  I believe Taylor often assumes the timing set at optimal - even his experiments with timing changes were done with variances from the optimum.  It can be demonstrated that, at lower power settings (like 65%), the same power can be developed at a more optimal timing (different than the compromise fixed timing, i.e. earlier than 25 DBTDC) at a lower fuel consumption (or, more power developed at the same fuel flow).