Sure, the description is important, and so is the reference. It's to explain why you think the user is a scammer or very untrustworthy.
Negative means:
You were scammed or you strongly believe that this person is a scammer.
We shouldn't use it for whatever else just because we explain so on the description.
That's just a leftover software artifact from the failing design of the system. It's a system of
trust. Take a look at the negative ratings I gave the bcash folk spewing "Bitcoin Cash is the real Bitcoin". Are they scammers? Well

Those ratings (or type of ratings) were approved by theymos.
Edit: The same way we shouldn't leave negative trust because someone edited a post of ours with a "FTFY", even if we explain so in the description of the rating
(Lauda will get this

)
Oh, yes.

Strongly disagreed. We do need a solution, developing another system is an option. Another way would be to reach an agreement with signature managers.
Re latter part: How do you do that? What happens in 2 our of 30 reject the agreement? How to prevent new managers popping up that don't accept the agreement? How do you enforce this agreement? While the suggestion may seem simple, it is most certainly not.
And because we can't/shouldn't leave negative trust to spammers then we need new ideas.
I wouldn't put 'can't' there. We most certainly
can, but apparently we shouldn't (better put, unfortunately we shouldn't).