Well yeah, but what I find surprising is he seems competent but he didn't mention the obvious: if there is an audible difference that is correlated with a measurement that's a whooping number of dBs down, so low it would be ridiculous to pretend to be able to hear a signal that many dBs down, then it means either the audible difference is an illusion, or the wrong measurement is being used... Maybe it is measuring something vaguely related to the actual cause and showing a tiny blip, but the actual stuff is much larger, at audible levels, and the measurement in question is not designed to catch it, so it does not catch it.
For example FFT does a lot of averaging so it is very good at catching harmonics, but if you have intermittent nonperiodic random spikes (think of vinyl cracks for example) or many other kinds of error and distortions that are not harmonics of the signal, not periodic, and/or have a somewhat random character, then it will just show up as a rise in the noise floor which is difficult to interpret. The energy of this type of error signal is spread over many FFT bins so it looks tiny, and it's not possible to know its real amplitude by looking at the FFT. So FFT is unsuitable for catching this kind of defect. People call it "noise floor modulation" because that's what it looks like, but who knows what's in that noise.