800 Score.com

This really isn't true - the algorithm is not complicated, it's just based on probability theory, and there's nothing proprietary about a 3-parameter IRT algorithm, which is what the GMAT uses. Certain details are proprietary, but those details are largely inconsequential; the fundamental foundation of the algorithm was published in academic journals decades ago.

The reasons company tests are almost always a very poor approximation of the real thing has little to do with the scoring algorithm those tests are using. There are two far more important factors: most important of all, company test questions are often a poor replica of real test questions. That's especially noticeable in Verbal questions on most company tests. Because company questions are different from real ones, they're testing different abilities than the real test does, so while scores might be somewhat correlated with real scores some of the time, they will be distant from real scores for many test takers. The same is true if you compare, say, high school calculus test grades with GMAT Quant scores. They'll be modestly correlated, but they're testing completely different things, and a lot of people will be much better at one than at the other. Companies also often haven't properly calibrated they're questions (assessed the difficulty levels of questions correctly), so often they're just doing guesswork when they produce scores. The real test is based on very carefully calibrated data for each question.

In general, I don't think test takers should put much stock in prep company test scores, no matter the company. Some are better than others of course, but none are a substitute for the official GMATPrep tests.