Practical Roman alphabets do need to stick as close to ASCII as possible. Particularly before computerised typography, getting hold of letters outside the Latin-1 and Latin-2 repertoire (letters and standard diacritics) was painful, and you’d avoid it if you could.
So if you had a choice between
tʰiantɕʰi pu xao
Tianqi bu hao
… well, really, that’s not much of a choice at all, is it. Practicality is going to overrule the universality of the IPA, by far: once everyone agrees that <q> corresponds to /tɕʰ/, there’s no reason you have to stock up on those extra odd letters again. Linguists working with Chinese can certainly remember that much.
There was fine print in the history of Pinyin, involving previous transliterations and the initial attempt at a Cyrillic based transcription; but really, this was an issue of practicality, no less than Albanian picking <x> and <xh> for /dz/ and /dʒ/.
In fact, the only practical orthography that in any way significantly depends on the IPA is the Africa Alphabet and its successor the African reference alphabet, which is used for several African languages. And that involved inventing uppercase versions of a lot of IPA letters, because the IPA had never been used in a practical as opposed to scholarly function before 1928. Hence:
Ɓ Ɖ Ɛ Ǝ Ƒ Ɣ Ŋ Ɔ Ʃ Ʋ Ʒ
So yes, there is a capital schwa and a capital esh. Who knew!