Problems with non-English hashes

Hello everyone and happy new year!

Currently I'm working on a pretty large list (3.9M hashes, trimmed down to 2.2M unique hashes) made of plain, un-salted MD5 hashes.
I know for sure that all the plains were created from Turkish people.
The problem is... I'm recovering very few of them (just a little less than 200 words).
Honestly, that's pretty a bummer: I'm far away from being a top cracker, but honestly with such a large collection I hoped I could retrieve a lot more of them before starting to have problems.
At the beginning I thought I wrongly detected the hash type or trying to crack something meaningless (ie token reset or something like that), but then some plains started to pop out, so those hashes are indeed passwords.

This is what I tried so far:

• Plain run of rockyou wordlist + all available stock rules
• Plain run of hashes.org worlist + all available stock rules
• Plain run of several different wordlists of previous leaks + all available stock rules
• Download the Turkish dictionary + all available stock rules
• Combining Turkish dictionary with wordlists
• Download list of names and celebrities + all available stock rules
• Started using masked attack (at the moment processing the rockyou-3 mask file)
• Replacing default charset with Turkish one (lower and upper chars)
• Started generating random rules
  

Sadly, I got very little improvement: only 2-3 passwords every time. However I noticed that there aren't any chars coming from the Turkish alphabet (ğ, ş ...), so my question is: am I doing it right?
Am I missing anything like file encoding or something else?
Maybe the originally chars were not encoded properly?

Any suggestions?