Acme treats all text as UTF-8. If the input text was ms-kanji, it won't be UTF-8 and when acme reads it, it will end up full of encoding errors - represented in UTF-8. Running that UTF-8 text back through tcs -f ms-kanji will produce gibberish.
You need to use tcs on the raw files before putting them into the editor (or almost any other Plan 9 tool). -rob
