The variance on adaptive tests is lower than on paper tests of the same length, so any variance you see in real GMAT scores is not because of the algorithm. But most of the tests you've taken aren't real GMATs -- they're company imitations of the GMAT. Some companies do a decent job of replicating the real test, and some don't, but no company can replicate the real test exactly. If you want to get a better assessment of your level, take another GMATPrep test or two, and average your scores. If you perform the same way on test day as you do on those GMATPrep tests, you can be confident you'll get a similar score. Your one GMAC test result is outstanding, so you should do extremely well on the real thing. Good luck!