Amp and speaker interaction cannot be simulated in lab amplifier measurements. Most all loudspeakers vary dramatically in load to an amp. Testing the amp and speaker separately is not an accurate measurement of how the components work together. Some speculation can be made sometimes but not always. So the method used is measuring with a simulated 4 ohm load and a simulated 8 ohm load but that is not how the system is operating.
My EE circuit skills at this level are not superb, but my expectation would be this is a factor of the stability of the amplifier which could be impacted by the impedance of the load compared to a pure resistance. I would expect amplifier designers such as Atmasphere consider this, and also that for most standard speakers (not electrostatic for instance), they expected range of speaker reactance is not so great they cannot account for it. When I was researching speaker impedance after discovering the high resistance cable, I noted that the phase angle seemed to be bounded though my research was not extensive. It seems an inherent element of most speakers is some significant series resistance.
This is a good point but not the nature of my question. If the amplifier has a characteristic output resistance that is significant, then knowing the characteristic input impedance of the speaker, we can simulate/model how that will change the frequency response. This has been validated by an EE strong in circuits so I am confident that is correct. I believe that would be dominant over distortion, but I am not fully confident in that belief. I do understand that high output resistance would also impact woofer movement which may not be easily modelled.
