But, then you lose the whole point of MBTI, which is to categorize into X number of categories.
Some things are clear cut in terms of it being a categorical data - male or female. Other things, it depends on the premise you start off with, so MBTI's premise is to group using X # of axis (dichotomous), i.e., categorize, thus it's kind of moot trying to find individuality through a test that's saying they want to group.
But, this is an issue inherent in any type of categorization. However, we can see aggregates of results and how it shows a normal distribution within each category. If everyone was falling exactly at 100%, you should raise a few eyebrows at the scale.......because it shows a clear bias. Variance is good. Hence, a good scale should give a range of responses, from 0 to 100 and thus, there will be those weirdos who fall around the 50% mark. It's a good thing for a scale to do this.
People are really different, and there's not a categorical scale that can ever hope to account for all the differences we see, because we'll be increasing the categories towards infinity, accounting for all the variables to measure differences.
MBTI has chosen 8 variables, to use to group people, of course, 8 variables cannot account for all human differences, nor should one fault it for doing so, because they're missing the premise of MBTI.
I guess, the best thing to do would be that the person is aware of how close they fall between an axis (near 50%), rather than take on the subsequent grouping to be absolute for them (i.e., take on the dichotomous reality and end it there).