Japanese support is only available for KEPUB, at least that's my experience. Even then there are selection problems for example if you have ruby (furigana) it can get in the way if it's per character instead of whole word. For the Harry Potter books I ended up post-processing them (merging adjacent ruby tags) to work around this problem.
Then you still have to select word boundaries manually and like as not it will delect instead of moving those annoying pin thingies and you can't really see what you're selecting because your finger is on it. That gets annoying very quick and you're back to reading in the webbrowser with rikaichan as a simple mouseover dictionary
