The first question this begs is: "is this really audible?"

That would be my guess too. In cases where there is a verifiable audible difference, beyond level matching, you probably have a tone control, possibly with some deliberate crossfeed for "ambience".

As I've previously mentioned, in the studies of loudspeaker preference, then measured performance correlates well with listener preference (except when the subject has some hearing loss and is looking for a prosthesis )