On one hand I agree that there are far too many frivolous ratings being sent. Ratings based on opinions or because of arguments, because of a clash of personalities, because of differing idea or views, trolling, and so forth, are both inappropriate and counter productive. They cheapen the entire point of the trust system, and serve more as a punishment against the person rather than a useful indicator of trustworthiness.
However, as Loyce has pointed out above, there are plenty of examples of accounts being correctly red tagged without yet having stolen anything or violated any contracts. If your entire ICO is plagiarized, then you are not a legitimate project. If you are advertising impossible ROIs, then you are not a legitimate project. If you are asking for users to enter their seed, enter their private keys, deposit before they are allowed to withdraw, and so on, then you are a scammer. I disagree that we should be safelyletting these users freely peddle their scams when we have the ability to pre-emptively tag them.
That would safety come under directly financially dangerous behaviors. Those would be type 1 flag. So long as the behavior is directly opening people up to being scammed or they look to be setting up other members to be directly financially vulnerable then they would get a type 1 flag.
It is the completely frivolous red tags that are for drinking lemonade or having a different opinion, or anything that results from a personality clash.
There would still exist a sensible credible and valuable warning for people that directly and clearly look to be attempting or setting up a scam.
We have to weigh that against the problems such a move will resolve.