These tests are all based on statistics, and so some people may fall between categories on different tests. Or the program chooses the closest response to your input. In Psych, we did the probability of these types of tests and though all claim accuracy, it is an 85% accuracy usually, so that's why they don't allow some things in court to hold definitive merit as they did at one time.
Also you could take the same tests with questions and answers which are similar but not the same or just worded differently and you are likely to answer oppositely.