[darkmighty]

How can you say the this approach is more efficient than one which keeps the engine at a constant load and varies propeller speed, for example? (it involves complicated fluid mechanics to figure it out)

This is a reason mathematicians loathe the overuse of "optimal" by engineers: in the real world there are often too much variables.

[lutorm]

You can't have constant load -- the power needed is dictated by the conditions. But what is known is that you want to operate at full throttle to be the most efficient, so the way to vary power is to vary RPM. (Propellers are also generally more efficient at low RPM, but I don't think that's what was referred to here.)