Currently, auto-completion does not work completely when the target language is Japanese.
It only works at the beginning of the segment, after punctuation marks, and so on (in short, rarely), possibly because words are not separated by a space or other marker.
So, I turn it off when working on an English-Japanese project.
Another reason why I turn if off, is that some kind of Japanese IME is mandatory for typing in Japanese characters, and it also has a word suggestion feature (predictive typing): the result is overlapping pop-ups, as shown below (purple one is from CT):
Some may say that it might be best that auto-completion works anywhere in the target segment; but, I'm skeptical about such possibility, and I don't seek it now because it does not resolve this distracting overlap after all.
So, to overcome the two issues of partial functioning and panel overlapping and to make the most of auto-completion, the best solution seems to be:
When I type in one or more "English" (source) characters, target terms will pop up for selection: some kind of adaptation of the current F2 list for auto-completion, like this:
Note: When an English character is typed, the Japanese IME is not activated.
I'm sure this solves both issues at the same time and benefits Japanese users incredibly.