If you are picking B 100% of the time blind as a preference, then you would also be able to correctly identify X each time. Not sure you understand how an ABX test works.
That being said, I still think the same thing could happen in a more rigorously conducted A/B and ABX test comparison. Due to the rather fickle nature of the brain, there is a big difference between choosing the one you like, enjoy more, etc. and determining which is which, as one does in an ABX test. I'm not sure if anyone has ever conducted such a study, but I believe that one really could net results where someone may pick "A" as the one that sounded/tasted/felt the best in an AB test, but would come up close to 50/50 in an ABX test where they try to determine which one 'X' is when using the same source material.
I'm reflecting on a study that was done with a dozen or so different jams in which they conducted taste tests between a large group of people to determine their favorites. In the end, there were two or three jams that garnered most of the votes for favorites and the list closely mirrored the results in which food critics ranked the best jams. They then did the test differently and asked the participants to actually analyze and describe "what" they liked about each jam and rank their favorites. As soon as they were asked to analyze and describe, the results pretty much became random and and the previous number one actually went close to the bottom of the list. Trying to describe what you like or determining which sample X matches is a much different state of mind - perhaps more like taking a test - as opposed to the gut hunch/intuitive sensibility involved in enoying a taste or savoring a sound and deciding which one you "like" more.
I'm not dismissing DBT tests - or even trying to discredit ABX tests, but do feel that there the fickle nature of the brain and the way it works prevent them from being as absolute as many would like to believe.