I added a "feature" to the Crimean Tatar transliteration that blocked the transliteration of tokens that had non–Crimean Tatar characters in them (like the name "Stewart"—there's no transliteration for "w", so it would transliterate as "Стеwарт", with the w still in the middle of the word). Similar situations exist for Cyrillic-to-Latin transliteration for Russian words with non–Crimean Tatar Cyrillic characters.
This was intended to minimize the need to use language converter–blocking -{markup}-, but it backfired here, and it's probably too hard to block in the general case, so it should be removed.