EDIT: ok, I've looked at the y-axis, we're talking about a max of 1.5dB range of difference at 20kHz, 0.5dB difference at 10kHz, and about 1.25dB of difference at 15kHz - ok they're small differences, but I suppose not ideal.
Honestly, I am also not understood how this amplifier could get a recommendation. Know about LS, but can not assess the impact of the amplifier.
When looking at the impedance response of randomly selected inexpensive bookshelf speakers on
spinorama.org (measured by Amir or Erin), each of these speakers will probably cause ripples in the frequency response.
Up to 65 ohms impedance due to the BR Helmholtz resonance and 30 ohms at the crossover frequency. But there are also LS that reach up to 80 ohms at the crossover frequency.
Amplifiers are not my specialty, is the correlation linear? So if at 4 ohm load difference at 2kHz the FR is changed by 0.2dB, will it be changed by 2dB at 40 ohm difference (or must the phase also be considered)?
Or asked another way, based on the 4 and 8 ohm load impedance measurements of the amplifier, can one simulate the effects of an arbitrary speaker impedance on the frequency response?
If the impact on the frequency response from 20-20000 Hz would remain below 0.5 dB (actually, one would have to consider the Q of the deviation as well, but it makes the consideration even more complicated), the amplifier is almost transparent, otherwise it additionally acts as an EQ and should not be recommended.
For me as an amplifier layman,
these analyses by Amir were much easier to understand: