What's wrong is that one feature gets one score, and another feature gets a much higher score. Scores are not based on the number of supported features, but on higher scores for certain features.