As far as I understand, the user in the first post suggests that you use a feature that allows you to teach the program how to parse you text. I found this: Training User Patterns | Finereader . Probably, that's what he was talking about. The help says this feature was (originally) intended for recognition of poorly scanned texts or texts with many unusual symbols.
The also says it usually takes no more than a half of a page to get it running - at least, if you got good scans. And, just as he said, the help also advises to use symbol combinations or character chart to input any characters missing from the "alphabet" of the input language.