You are right. As I said, the Zcomp circuit must be developed for the particular speaker, since the hump center frequency, Q, and max Z is specific to the speaker model.
An ideal amp has zero output impedance, and is immune to any non-zero speaker load impedance. So the speaker SPL curve does not become skewed, in spite of a non-flat impedance curve.
Take the non-flat true impedance curve, and a non-ideal amp, and the impedance shape "creeps into" the SPL (amplitude v. frequency) response.
For SS amps, especially those with high current capability, the effect is quite small. So Zcomp is not really necessary. Although, even if it is used, the swinging phase angle will be flattened, just as the impedance amplitude is flattened. In this case, the SPL curve remains unchanged, in spite of the flattened amplitude and phase impedance curve. Even the SS amp appreciates a flatter load.
With high output impedance tube amp, let us say the SPL curve becomes *uncalibrated*, and requires *EQ* to make it as flat as if the speaker were driven by SS power. Rather than take a brute force approach of EQ, flattening the impedance curve via a Zcomp circuit will do the trick, and the tube amp will certainly like a flatter Z load.
It is a simple matter for my measurement program to not only calculate the phase curve, but it can also display the real (resistive) and imaginary (reactive) parts of the impedance. When the real part is kept flat and high in value, the amplifier is *happiest*. Consider that a nominal 8ohm speaker could still have a real part of 1 ohm in a narrow band, where the impedance changes dramatically.