This post for example was 9.8, so I figure the sample size will determine whether some people score high or low, as their posts will vary widely.
That's a problem I didn't want to deal with though, but it is pretty big determiner - in what mood are folk speaking in? passionately? angrily? irkedly? semitrollingly? It changes the scores very much, but I did try to get at least 3 good paragraphes out of all these folk though. More than 100 words were analysed for everyone.
I have a suggestion: Perform the analysis on a per-comment basis. Select say 10-20 comments 'randomly' (doesn't have to be perfect, but just try not to pick them with a bias for long or short posts) for each person, measure the score for each post, and average the score for each poster's posts. This will give you a more-standardized 'score per post' rating for each poster. It will certainly take a bit more work to perform, but it will give a better 'big picture' look at things.