:-) Sounds like your problem is not the conversion, but validation of the new file. I am not sure how anyone who isn't familiar with your dataset can help. If some one else do the conversion in a different method, he will still be susceptible to the same mistakes (or other mistakes).
Maybe a better course of action is that you should validate the converted data, Maybe by comparing for some observations in the original db or thinking about axioms you know should hold in this data and see if they are not skewed....