Isn't there going to be any change in the efficiency of the generator?
Perhaps trapezoidal magnets with trapezoidal coils will be more efficient or not?
And an emf that is not sinusoidal doesn't imply that there will be harmonics that are useless and will only produce losses?
What do you expect with efficiency. If you are generating sine waves for a conventional load then yes harmonics will lower efficiency.
Once you decide to rectify with simple diode rectifiers then the harmonic content must go up and the power factor will be less than unity.this is true of rectifiers feeding resistors. Once you clamp the voltage to a battery then harmonics are inevitable and you will suffer losses, why worry about a few % harmonic distortion in the open circuit waveform when the loaded waveform is chopped beyond recognition.
Another factor is that if you could achieve 100% alternator efficiency you would have a constant speed machine that would produce a dreadful match to the prop power curve and you would be an order of magnitude worse off overall than accepting an alternator efficiency below 60%.
If you want absolute perfection then go for mppt and use a low distortion active rectifier such as a Vienna rectifier. You will produce an excellent machine but it will be impossibly non cost effective. Fine if it is your pet project done for love, but not a commercially viable scheme.
Just to add a little more to your problems, using trapezoidal magnets and coils will not guarantee any approximation to a sine wave, the inherent flux linkage is trapezoidal ( near rectangular so to get a sine wave from phase volts you will need to proportion the winding distribution so that the odd harmonics are mainly suppressed.
With normal proportions the phase volts of most windings is near triangular with in phase 3 rd harmonic. The line voltage of a star winding will be very close to sine wave but this will only apply into resistive or at least linear loads. All gores out the window with any form of non linear rectifier load.
Flux