I am experiencing the same problem. 6.03 worked fine, but 6.12 doesn't. It's as if mIRC isn't doing the JIS/SJIS conversion at all - whenever other people type in Japanese, I see the same kind of garbage characters that I used to with older versions without JIS/SJIS conversion.
As I upgraded from 6.03 directly to 6.12, I am unable to check whether the bug started occurring in 6.1 or 6.11. Sorry. Anyway, the bug is there.
Warning: long Unicode rant follows.
I've never been able to actually write Japanese in mIRC - if I try to do it, mIRC just receives and outputs question marks. This is a sign of the application not handling Unicode text input properly.. I suppose it's because the default language for non-Unicode programs is Finnish in my case, and because Windows can't convert Japanese characters to anything in the Finnish charset, it will try to feed Unicode characters to the program.
I don't expect mIRC to ever support all the different character encodings for text input, but it'd be nice to be able to at least choose between "system default" (which is currently always used) and "utf-8". UTF-8 text isn't rendered properly either. I know I wouldn't be able to type JIS/SJIS properly even if mIRC supported UTF-8 translation, but at least I would be able to type Japanese text to other people whose clients support UTF-8..
The option should be on a per-window basis, since most European IRC channels are still stuck with country-specific character encodings. But I'd love to be able to use Unicode and UTF-8 in specific channels and/or chats.
It really shouldn't be too difficult to implement, because I managed to write a simple Windows GUI tcp/ip chat application in C using Unicode in four hours. The only thing that's needed is just to use the Unicode versions of WinAPI's class register and window creating functions (so that they receive messages in Unicode), write a UTF-8 <-> UCS-2 translator (since the messages received from WinAPI will be in UCS-2, but UTF-8 should be used in IRC) and change the text renderer to output Unicode (shouldn't be hard, since it already supports multi-byte characters). And add a Unicode option to the menu of each channel/chat window.
I wish the answer was a standard open source answer ("The source is available, go and fix it yourself"), because if it was, I'd do it in an instant. :-)
A partial solution is to use Putty (Windows SSH client) and irssi (*nix IRC client). You can already chat in IRC with UTF-8 that way. But I kinda like mIRC..