While curves like those can provide some insight into the characteristics of a particular amplifier that is being reviewed, I'm not sure that they help to answer David's question.
As I interpreted it he was asking if optimal power RATINGS could be defined for different basic categories of amplifier design. Things like tube vs. solid state, the type and quantity of output tube in the case of tube amplifiers, push-pull vs. single-ended, class of operation (A, AB, D), balanced vs. unbalanced, amount of feedback, and other differences in circuit topology.
The first problem is that curves depicting distortion or distortion + noise as a function of power level are not much help in trying to choose an optimal power RATING for a given application. Unless the user listens almost exclusively to very over-compressed recordings, the amplifier will be operating at points pretty much all over those curves, as a function of the music.
And in comparing performance at different power levels, a given increase or decrease in THD or THD+N may be subjectively either better or worse, depending on how the THD is distributed among different harmonics, on the relative amounts of THD vs. noise, and on the spectral characteristics of the noise.
Also, with many amplifiers, especially those using significant amounts of feedback, the THD numbers will be negligible even on the higher parts of the curve (well away from the bottom of the "U"). While TIM (transient intermodulation distortion), which is worsened by feedback and which is not normally specified or measured, may be a much more significant contributor to subjective sound quality (depending in part on how much propagation delay there is through the amplifier, that being another variable that is generally unknown).
So I don't see any reason to expect the amplifier to necessarily sound best when operated at the bottom part of the curve, and even if it does that would say little or nothing about what power RATING would be optimal.
Best regards,
-- Al