That's probably impossible for anyone here to answer.

Besides the relevancy of the 'hit rate' that you described above, there are two other factors that are equally as important to your score on the actual GMAT:

1. Your average time per question
2. The difficulty level of the questions you're getting right and wrong

In other words, on the GMAT--if you're answering 15/20 quant questions correctly, but the 5 that you miss are easy in difficulty then you will have a much lower score than if the 5 you missed are considered very difficult. Moreover, even if you can answer all of the questions in the OG correctly, it does not mean a whole lot unless you're averaging ~2 minutes per problem.

Therefore, your 680 & 690 on the GMAT prep software is probably the best estimation for how you'll do on the actual GMAT. And also don't forget that you may be more stressed on test day than you are in the comfort of your own home taking the practice CATS. So from my own experience, I would actually suggest that you're score may be slightly lower on the actual GMAT than on your practice CATs

I agree with you. Since th difficulty level of OG questions can not be fairly known, the accuracy rate might need deeper introspection or such an activity may turn even useless. Timing is OK for me. I am very will with in the limits. As yo said, the actual score may be lower than my test scores. Let me work more. Thanks for your comments.