Meta-analysis: Decent personality tests can be modeled as providing an interesting question, some common answers to that question, and some observations about what those answers usually imply. For example, in the case of the love languages test, the question is 'what kinds of displays of affection do you notice most easily?' According to the first sentence of its wikipedia entry, the Myers-Briggs test is probably asking 'which biases are you most prone to?' (This may not be its actual question; I'm not very familiar with it.)

If none of the categories seems to fit you, it could be that you answer the question that the test asks in a way that the test's creator didn't predict, or that the question itself is not a useful one to ask of yourself. In the former case, you can still learn about yourself by figuring out how you do answer the question. In the latter case, it's probably best to ignore that test altogether, though there is also value in the pattern of which questions are or are not useful and important to you, and why.