Frankly, I don't trust your judgments of edit quality as a new reviewer on Stack Overflow. You just unlocked the review privilege earlier today, and have only reviewed 20 suggestions at the time of this post. That's just barely enough to make the accuracy of your percentages algebraically feasible. What makes you think the suggestions you've reviewed today are representative of edits in general?
Your reasoning regarding the benefits of your proposal is similarly suspicious.
1) More positive edits will be accepted, improving the global formatting of the questions.
Are you suggesting that many positive edits are rejected because reviewers don't want to award 2 rep? Regardless, diminishing the reliability of that incentive would diminish the frequency of suggestions somewhat, which would mean fewer improvements.
2) "Rep whores" who try to edit lots of question won't do that so often unless the edit has more value.
"Rep whores" aren't necessarily submitting worse edits. As above, this conflicts with your first point.
3) The edit decisions are faster for the reviewers. Most of the edits are of category 3.
You're suggesting we give reviewers an extra option to consider. How would this make it faster? I have a much easier time judging whether the suggestion is valid than I would if I had to judge its rep-worthiness on top of this.
4) Editors and reviewers will probably be happier to see that the site improves overall (because more edits are approved).
Redundant with your first point, and suspicious for the same reasons.
Another issue to consider is the value of editing as a viable pathway to unlocking privileges. Gaining reputation might be relatively trivial on Stack Overflow if you know what you're doing, but it comes much slower on some sites. Suggesting edits is a nice alternative to asking or answering for users who want to help build communities. It's particularly useful when one doesn't have the expertise to answer many questions and doesn't have any particularly original questions to ask. Editing involves a lot of reading, which builds that missing expertise over time and generates questions.
In my own case, I've become a much better statistician by editing my way to 2K rep on Cross Validated. In the process of looking for questions to edit, I've found many I could answer that I wouldn't have looked at otherwise, and came up with my only question thus far. Now that I have the badge too, this doesn't happen as often on Cross Validated.
For the moment, I've moved on to Meta Stack Exchange and Travel Answers instead! Edits aside, I've made most of my contributions on each while looking for posts to edit. My contributions are definitely better for all the editing I've done. I've gained a much better understanding of the site-specific cultures than I began with. As your question indicates inadvertently, you should not underestimate the complexity of this acculturation process. One thing I've noticed here: feature-requests for changes to the rep system seem particularly hazardous for your rep score. Is that ironic? I can't decide...Guess I'd better go edit some questions on English Language & Usage next.