This can be the case, yes - the # correct will not necessarily give you much of a clue about the score (beyond obvious cases, such as everything right or everything wrong :)
I don't know whether the GMATPrep algorithm does the same thing that the real test does - that is, some number of the questions are experimental and don't actually count towards your score. If GMATPrep also does this, then that could explain why you seemed to get so many right but still have about the same score in the end: your performance on the mix of counted to non-counted questions could have been skewed.
The percentages that you gave - that's the percentile ranking that you earned? So you scored 49 on the first one and...48 on the second? (that percentile ranking doesn't match the current scale, but maybe you downloaded GMATPrep a while ago.)
If you're scoring in the high 40s, then you are doing very well, so nice work. I'd focus more on the individual questions: how can you learn to get better? (That is, if you want to lift your score more. Your quant score is quite high already, so possibly verbal needs more attention?)
This can help you learn from those questions:
http://tinyurl.com/2ndlevelofgmat