I love a good irony, and we people create them a'plenty.
I think there actually is a general way to distinguish between confirmation bias and a conclusion based on analysis. It has to do with recognizing and expressing the nature of the system. MBTI is by its nature an approximate system. Confirmation bias will use the natural flexibility and indistinctness of the system to mold it into preconceived hard, absolute results. The analytical mind will form conclusions that are a natural outgrowth of the system. MBTI is most useful when the fuzziness of its boundaries are kept clear in the mind and are integral to any conclusions based on its principles.
I actually don't think MBTI is very fuzzy (or at least it isn't applied that way), which is essentially my problem with it. The only room for fuzziness in the system is how the four functions interact with each other, not within the functions. (I guess this is a pretty nit-picky point; forgive me. It seems we agree anyway.)
As long as the fuzziness is clear (heh, that sounds funny), you're right, it's possible to use the system correctly. Again, and this is really the point of the thread, to actually use it this way seems to take as much work as building a framework from scratch for each person. And I find that I actually do that. MBTI might give me a tiny jumping off point, but not much.