Well I definitely disagree that anything you're describing is curving. Curving by definition involves comparing students to each other or ranking students.
I agree that ONE of the ways they could assign scores is by using the model you describe and deciding that a set amount of points (which would be more likely to be somewhere around 50 than 75 if our total is 100) is required to “pass”. I actually posted an extended scenario like that about two weeks ago. Short version is, for instance, testlet 1 questions are worth 2 points, testlet 2 and 3 if you do well are worth 5 points each, sims are worth all different amounts (based on a difficulty score assigned to them by the number of correct responses when they were pretest), and you must accumulate 275 points to pass. A med-hard-med test might only have a total possible of 350 points, a med-hard-hard test might have a total possible of 440 points. I actually think that's a better method.
My original point was that we can speculate, but they have not given us details about how the score we earn is mapped to the score we see. That doesn't necessarily imply curving, and does not exclude the weighting of questions. As a matter of fact, both methods could be combined. The scores resulting from the scenario in the above paragraph could then be ranked and assigned a percentile.
The reason I lean towards a percentile coming into play somewhere is because of two things that are unexplained by the method in the paragraph above. One, high scores are 99. Never 100. Classic percentile scoring. The second thing I just can't reconcile with the “if you get this amount of points, you pass, period. If you don't, you don't.” is that we can't leave the test center knowing our score. A lot of people have complained that they left the testing center with their GMAT or GRE score but you have to wait for the CPA score. That's not quite as convincing, but it's still a factor.
AUD - 88
FAR - 90
REG - 85
BEC - 88
.