Oli, the problem arises when you award people a rating on 90, 90%, RR90, etc., and classify this as a good thing when the reality is that they average an NMR per game. 90% is a good grade on a literature paper, or a calculus exam, but it's horrible for Diplomacy. I think it's doing a disservice to everyone to call 90% a "good" rating. Your suggestion of removing the attached language is a good one, but it's disappointing to know that you think that 90% performance is acceptable behavior.
I think Retillion put it best a few pages back, or perhaps on the other thread, where he started that if your wife is faithful to you 90% of the time, that's not a good score. And the point of this exercise I would hope is to provide incentive to players to improve their reliability so that all players in the community benefit. Once again there seems to be a driving force here to assure that the mediocre, the lazy, and the spiteful players go unnoticed, unpunished, and worse, encouraged to continue behaving this way. If you ignore this element in the ratings, then no matter how clever you make the math and the scores, and the statistics, you're still just putting lipstick on a pig.