When a Garmin map is created with a specific character set (typically a character set based on the Latin alphabet), names must be “translated” into that character set. That is no problem for English names, and most other languages written in a Latin-based alphabet. In other cases, an English version of the name may be provided in the data (e.g. name:en=Tokyo), then that can be used.
In all other cases, some “translation engine” is used. For classical Chinese characters (i.e. also Kanji), it uses the Pinyin transliteration.
Please note that the problem is not with Japanese only. Also Indian alphabets (including Thai) show issues of that translation process (there it is wrong order of vocals relative to the consonant).