In my experience, the Princeton Review CATs have been somewhat inaccurate. From what I have read, they focus HEAVILY on the first 10 questions of each section. This jives with what I was taught in my Princeton Review Classroom course, where they emphasized that, essentially, the only questions that really matter are the first 10.
This couldn't be farther from the truth, as anyone who has taken the real GMAT will tell you. Go back and look at your results from the PR CAT. You will probably see that you got 1-2 wrong in the first 10 of the math portion. By doing so, you basically guarantee that you can't recover and score >700. I have taken multiple PR GMAT CATs (scores listed below in my sig), and usually missed 5-10 math questions. Conversely, on my
MGMAT and GMATPrep CATs, I have missed as many as 16 on math and scored in the mid 700s overall.
I would recommend using the Princeton CATs just for practice with pacing and timing. And yes, the math on PR CATs is a joke. Just take a
MGMAT CAT and you'll get your socks knocked off (but still actually score well because their scoring algorithim is much more realistic). Ultimately, the GMATPrep CATs are said to be the best predictors of your score, as they are the ones actually made by the GMAC itself.