Given how the test was administered - it was a yes/no "do you know this word" test with some fake words mixed in to discourage people from saying yes to words they don't know - "know" is a bit of a stretch here - "recognize as a word" is probably the better description of these percentages. I don't think the results are all that meaningful because none of these words are used much - most of them are used only in contexts that already imply an uneven gender ratio. On the other hand, there are probably a lot of common words used in casual contexts that are recognized by virtually everyone whose usage is significantly skewed by gender - those are probably much useful for identifying gender.