Perhaps I've misunderstood, but I think it's not the text that is unicode complete, it's the transliteration rules that are.

Unless you install special rules you won't be doing extra computation.