The real problem here is the lack of data that we have to work with. Using a copy of your old database and comparing some of the new working entries might give you a test base that may lead to the creation of a "cleaner". Along with the questions @Lynne posted, I have questions like: is it a character-set encoding problem, a collation problem? Performing small tests will answer these questions. There is not going to be a simple answer forthright, unfortunately.