Um..... why exactly can't you measure a worst case test load?
It sounds to me a classic case of "The measurements are incomplete therefore the listening is more relevant"
But for electronics I'm not sure that should ever be the case given the dubious nature of the industry. The measurements should always be weighed greater. If that means measuring into a 1 ohm load @ 75 degree phase angle for a $5000 amp @ 10hz to 22khz, I don't see why not. It's pretty insane but the measurements should always represent a worst case scenario not a middle of the road scenario.
Likewise with loudspeaker measurements. Why only measure 30 degrees off axis when the sound at 60 or 70 or even 150 degrees off axis has been shown to affect perception?
I understand as a reviewer time constraints etc come into the picture but it's unfair to do half-measurements - and then claim that they don't help identify what you heard.
I'm not criticizing your measurements here Gene. They're great and appreciated as are the subjective commentary. But It's a bit disheartening to see you post the measurements and then claim they're inadequate relative to your experience. If your own speakers are a more complex load than the test signal, then as pretty much the publication leader we expect you to address that objectively rather than use the limitations of the measurements as evidence. I don't really like the idea that something that can be heard but can't be measured. That's awful close to "cable" territory to me. It can either be measured or you choose not to/lack the capability to measure it.
Sorry to rant especially since this was a really good review but your comment that you can hear things that you can't measure just doesn't sit well with me. I'll admit my measurement capability compared to yours is infinitesmal but I also don't make such claims. I know you know this but it feels you just word things in a certain way that gives teh wrong impression.
It seems like a very good topic for an AH article - what aspects of amplifiers when isolated - are actually audible in an ABX / single blind test? I have no doubt that if you can hear it, you can measure it but complexity is the keyword.