In fact, it converts ascii characters into "extended ascii" characters. So the bytes count is correct.

If you're interested in pack-to-unicode, I did it too: https://gist.github.com/xem/7086007