Sounds to me like your encoding handling was broken before (what you describe seems like UTF-8 that was interpreted as ISO-8859-1), and might be correct or possibly broken (but in another way) now.

Without samples of your inputs and outputs it’ll probably be impossible to guess what’s up.

In the meantime, you should read The Absolute Minimum Every Software Developer Absolutely, Positively Must Know About Unicode and Character Sets (No Excuses!).

Makeshifts last the longest.