If what you're asking is, 'Do people vary in their ability to differentiate between various mp3 rates and the original?', the answer is yes. Part of that is training. Part of that is that's just the way people are. Look at people like a statistical distribution that more or less looks like the following...
where the Y axis (up and down) represents the number of people and the X axis (horizontal) represents something else. Maybe it's bit rate. Maybe it's height. IQ tests. Some variable.
You might be interested to know that there are people who are such good taste testers that they can tell you if the cookie they're eating contains reprocessed material and from what factory it probably came from. It's all in the training.
So, it's not that ABX is flawed because some can't hear a difference. People are just different. That's why when developing things like codecs, you look to find your most sensitive listeners and they might not all be sensitive to the same thing. When they can't reliably hear it, you can be pretty sure most others are just whistling in the wind when they say they can.
There are software implementations of ABX where you can listen to something like the original vs. some mp3. IOW, with just you and the PC, you can implement a DBT. Myself, if I go compressed, it's 320 of course with the decreasing costs of storage, lossless becomes very attractive. So you see, you just can't generalize and say mp3's are always distinguishable from the original or that they can't be distinguished. There's more to an mp3 than the three characters that represent it.