While this does seem to be a glitch, it actually seems to be giving you more information.
What I'd assume happened is that the code that generates the email for potentials and the code that color codes the ratings are just slightly different. Most likely done by two different developers or one developer with a lot of time off between them.
So, for example, what you'd get is in the email, you have code essentially saying "if potential is greater than or equal to (>=) seven, call it average, else call it low" (ignoring the high, obviously). Then in the ratings, it probably says "if potential is greater than (>) seven, call it average, else call it low".
Now when you get a guy with seven points of improvement left in a category, it's going to show as average in the email and then low in the ratings. But it would tell you that it's right at the edge of the low, so you would know exactly how much improvement this guy has left rather than him coming in at average and you expecting anywhere from 7-20 points. Now you know he's only getting seven points.
It's an assumption, but very likely the case.