Yes, I think reference compensation is the missing piece of the IR puzzle. IRs are, in theory, extremely accurate. The problem is that your average power amp isn't perfect, especially when driving a complex load. This causes inaccuracies in the frequency response, especially around the resonant frequency.
IR compensation allows you to capture a reference IR of the voltage right at the speaker terminals and then apply the inverse of this to remove the errors.
Hands down, but i think i will disagree to a certain degree: If we`re talking of a "average tube amp" this make indeed a noticable difference. If we call a solid state amp, in my example a Matrix GT1000 or even more linear amps, like Brystons (i think FAS did use for shooting IRs in the past?!) "averaga power amps", an inversion to flatten out completely the power amps influence will do nearly nothing to the IR result - especially when comparing the influence of different mics and mic-positions.
I tried to discuss this in the past when Clark talks about this approach by using the tube power amp of his Mesa Amp for IR shooting:
http://forum.fractalaudio.com/axe-f...k-kent-new-irs-way-teaser-12.html#post1085114
Again: This shows the frequency response of the Matrix GT1000 at its speaker out, while the amp is seeing the load of a 2x12" Speaker cabinet (using a cheap Behringer GI100 DI Box to link through to the cabinet & an direct line-level up-link of the matrix):
The most influence (non-linearity) we`ll see between ca 1,5kHz and 4Khz (approx. loss of 1db...). IMHO, adjusting the mic half an inch different will make a mulit-times more change to the IR result, no?! I have compared selfmade IRs with the same IR+power amp inversion, the difference is only little audible in the bass department, apart from that nearly inaudible. But as you can see: the loss from up to 10db of the CF200 is not represent by the Matrix:
Note: This IR was made by a Hughes & Kettner CF200 in PA mode, which shows a much more non-lnearity frequency response, than the Matrix, especially in the inaccurate bass response!
If we talking accuracy, added informations how to tweak the speaker page for IRs of individual Speaker Cabinets would help a giant step better. This is and remain (because by design of a modeller) the missing piece of the puzzle, which can`t be closed automatically, but manually by knowing the behavior of an actual cabinet.
So, the missing piece could only be closed, if we get the needed speaker tab informations of the captured cabinets. IMHO.
Mics, Mic positions & Speaker tab informations influence the result drastic, a good solid state amp assumed, the tool for power amp inversion is a nice one, but with minimal influence to the result, IMHO.