I still haven't figured out a way to *reliably* extract Japanese SJIS zip files on Linux. unzip/7z work on files with extended UTF-8 filenames. unzip always mangles SJIS. LANG=C 7z passes through SJIS and convmv can convert to UTF-8, but that mangles UTF-8 extended zips.
I mean I'm fine specifying the encoding manually. Also, even just *not mangling* the bytes would be better, then I can just unconditionally use -UU (ignore UTF-8 filenames) and LANG=C and convmv.