For keyphrase density, we don't count synonyms because we want to make sure that the main keyphrase occurs a sufficient number of times in the text. We do count synonyms for distribution, however, because this check has the goal to make sure that all portions of the text reflect the general topic.

Taken each on their own, these two assessments make sense. Together, however, the general user experience is a bit counter-intuitive. Adding synonyms will change keyphrase distribution, but won't do anything to your keyphrase density. Additionally, when you don't have synonyms set, you might easily get a situation where you have a bad distribution and a good density; if you then want to improve your distribution, adding keyphrases might send your density over the recommended maximum.

We should think about how we can make these two assessments work together a bit better, e.g. by communicating better that synonyms might be the key for a good distribution.