Hi. The 2 algorithms are supposed to be quite close with only operational differences.
However, there are differences with the question banks (what questions you get to see) and also stress. In general the algorithm is pretty standard - we even built it for the
GMAT Club tests - it is quite complex but standard and it works fairly predictably too but the part that you won't know and won't be able to figure out is what question parameters were responsible for your score. you need to know difficulty of the question, its discrimination parameter, and its guessing parameter - we wont' find those out unfortunately for your real test or for the GMAT prep even.
The test builds a picture of you (and your score) and it is a confidence interval that then becomes a score and the test has to reconcile mistakes on easy questions and correct hard questions to figure out which score to give you. These two are conflicting.
Ultimately it is not the number of mistakes but the ability to consistently answer easy and medium questions (don't worry about hard) that gets you higher scores becuause if you have a solid base and even with a few hard ones correct the algorithm is more certain about your ability.