Sorry dude... say what you will about the method, but what I heard was not a difference that was as simple as a level mismatch or a sighted AB comparison. It wasn't anywhere near that close.
Yet it's not merely me 'saying what I will'.
Sighted, poorly level-matched methods are simply insufficient to support your speculations, much less prove your case. That's what scientific method 'says', not me. Proof requires actual measurements and bias controls to avoid the errors your methods are prone to.
DBT of hardware is hard. But you could at least try this (a measurement): digitize the outputs of the same musical sample, played by the 'different sounding' units. Compare frequency vs level of the resulting files. You can post them to me and I will do the analysis for you if you like.
'Rolled off' commonly means significantly tilted treble-to-bass balance in one presentation verses the other, and that is very measurable. If you see that in measurements, then you have a basis for investigating why it's happening.