I'm in agreement with Tebowwww, the 75 is the score they assign to whatever percentage correct 45% of the test takers got above, so it doesn't matter if it's a 75 or a 50, the same number of people would pass. That's an awkward sentence. There's a few ways they could do it.
If they grade your raw score (as opposed to the assigned score you see) on a percentage of points basis, they might look at where the 45% mark is, and let's say it's 61%. They mark this up to a 75 and adjust the 45% above that line to their adjusted score. It probably isn't an “add 14 points” method but a percentile method, which is why they can technically claim it isn't “curved”.
If they do it as they should since they use adaptive testing, there might be many different possibilities for the total points a candidate could end up with, but there's a minimum amount of points required to hit that 75. So for instance, if you get all the questions right on the first testlet, and you get a hard testlet for your second and you do well on it, so you get an especially hard third testlet (because they're experimenting on you and separating the very highest level of test takers), your total available points might be 325 where a person who got a medium, medium, medium testlet's total available points are 150. The examiners might decide that a candidate needs to get 200 points minimum to pass, and map the 200 to a 75. I doubt they do it this way though. It'd be complicated and I really don't think they care enough to do all the software writing that would be required.
If instead of “the passing score should be 70” you mean “the passing rate ought to be raised to 55% from 45%” then I guess that's a separate argument. I personally believe that the whole exam is deformed, and the process isn't designed to achieve the goal of protecting the public interest. I believe it would be better if there were a basic baseline of information that any accountant who practices needs to know and if 80% of the candidates know it they pass, and if only 10% know it they pass. Since we're not going to do that, I do believe there is some value in “curving” since it smooths out differences between testing windows and waves, and years for that matter, and is an automatic adjustment for bad or ambiguous questions.
AUD - 88
FAR - 90
REG - 85
BEC - 88