It is true that the self-selected (and lower) cadences have been shown to be more efficient, meaning that more of the body's total work is used to make the pedals go. And also, more of the leg work spins the crank (FE, force effectiveness).

Yet there is one thing that always concerns me about it, and I googled up the paper being discussed (The relationship between cadence, pedalling technique and gross efficiency in cycling) to check. It is that trained cyclists are likely to train most at those self-selected cadences, and as a result of that training one would expect them to have a better pedal stroke (FE) and have conditioned themselves to be more efficient (GE) at the trained cadence. That would also be consistent with elite riders, who train generally at higher cadences, being more efficient at those cadences.

So if my concern has merit, these differences in efficiency might be due in part or in full to training and are not a consistently applicable physiological fact.