eBay's rating system has always been flawed, and this is pretty much identical. What would really be needed is a scorecard, which would not really leave much room for subjective evaluation.
If someone flakes, you put em down for a flake. If you see someone with 10 flakes, it says something (like don't hold it for him if you have other buyers lined up). The guy's maybe not out to screw anyone, but has a hard time making his mind up.
For actual deals, fill in details (online survey style) and tabulate a final score. Was payment sent promptly? How long between payment and a tracking number? How long for the actual delivery? Was the item as described? etc. All the answers get a +/- value, all values are added up, and the result is a +/- score. For the sake of the example, say the two extreme scores are +5 and -5, you can count -1 to +1 are a neutral, and the two other ends as a + and -.
How does that help? Say that the guy is scoring slightly below average on all counts (e.g., the guy was slow shipping, but just a bit; the item had more wear than he'd claimed, but it's not broken or damaged, etc) . Most people would currently give a neutral or a positive. We tend to be forgiving that way. But from a totally objective point of view, that was a bad deal that deserves negative feedback.
And a longer comment line, where people can explain things, and maybe make a subjective observation if the final score doesn't quite reflect how they feel.