Thanks for your answers.
Chinese is of course not the sole language to have a romanisation (a.k.a., transliteration) scheme, since passports from any country must show Roman characters (along with the native language).
So, we could maybe add a '-Latn' subtag to the language code?
For instance, for many Ukrainian POR, the Ukrainian transliteration (uk-Latn) is not the same as the Russian one (ru-Latn).
Note that Wikipedia has got apparently the same issue. As far as I have seen, the language template (e.g., "{{lang-ISO|}}") does not contain references to romanization/transliteration, and some have found work-arounds, like in
http://en.wikipedia.org/wiki/Kozelshchyna_Raion . But I have not digged that much in Wikipedia, and they may have a cleaner solution for that issue.