Accuracy of official GMAT software?

I guess my questions was more such that you don't think my raw scores seem high given the large number of questions that I got wrong still?

No, it's normal. Once you reach the tough problems, they don't penalize you hard for every error.

On simulation, I got about 5-7 wrongs in each section and still scored 770.

That sounds about right. You can get a lot wrong on the quant and still score very high. I think for verbal, there's less room for incorrect answers.

As an additional anecdote, I got 15 wrong on GMATprep and (somehow!) scored a 48 in Quant. That is 4% below a 49, but maybe 3-4 questions make the difference? I don't know.

Does actual GMAT tell you anything afterwards about percentages of questions right, etc.? Anyone have any datapoint about what sort of percentages have led to certain scores if so?

No, it just gives you your score and percentile. Although most people will tell you that percentage correct doesn't matter, I tend to believe that it does. If you want 700+ you need to stay in the 10 wrong/section area. Any more and your score will dip under 700.

I tend to agree with you, the less wrong the better!

From GMATprep, I had 12 wrong in verbal for a 39, and 7 wrong for a 44. Missing 11-15 has given me between a 41-48 in Q. For scores ranging from 690-710. However, I usually can tell if I am doing well, the questions get harder in Q, while in V they stay the same. I usually get the first 10-11 correct in V, and miss 2-3 out of the first 9/10 in Q.

Keep in mind, there might be experimental questions in GMATprep that affects how the score breaks down (as in Powerprep).

The moral of the story, get as many right as you can! And don't miss the first ones... (I missed 10 of the last 12 and scored a Q48 in gmatprep).

Overall, the test is adaptave. This means it will be different for everyone depending on your answers. Also, I have a theory that a hard question at the end of the exam has a slightly less probability of being answered correctly due to time constraints. Not that it "counts" less, but the necessity for getting it correct at the end versus the begining or middle is less. This means, that the test could consider where you are in the test as well as your time left. I don't know, I am sure that it could if the programmers wanted it to. The less time you have the lower the prob of getting it right... I don't know, just a thought. This would mean that easy questions at the begining HAVE to be answered correctly because you have all the time in the world in theory.

That might be crazy, but kinda makes sense to me.