I think it's pretty clear that Olive's application for trained listeners is practical speaker design & evaluation, NOT "science". With regards to the latter I tend to agree that 'trained' can't be assured to give identical results to 'random' under all circumstances.That would also be my guess, i.e., that what the study concluded, at most, is that there is high likelihood that the trained listeners will notice the same things and prefer the same speakers as listeners who haven't had the training. But it does seem to me that some people were suggesting that the conclusion of the study goes beyond that. We'll never know exactly what they intended because to their way of thinking there's nothing wrong with simply pointing to some study and asserting without any elaboration that what another person had written is in contradiction with the conclusion of the study.
But personally as an experienced, (not necessarily "trained"), listener, I'd just as soon try a speaker deemed excellent by trained listeners as by average joes.