[randallsquared]

Adding some Chinese (from wikipedia) does actually show UTF-n as worst-case compared to UTF-16 and UTF-8, at least. I got UTF-16: 0%, UTF-8: 5%, and UTF-n: 7%.