Spoken words are constrained to match mouth movements in a different language, the written forms don't have this restriction.

I doubt they were considering language learning when subtitling, unfortunately.