There was issue #391 about 'underlying type' for a reference used in [over.match], fixed by #1013. But this misusage of 'underlying type' is still exists in "this document", in the definition of bind directly :

In all cases except the last (i.e., implicitly converting the initializer expression to the underlying type of the reference), the reference is said to bind directly to the initializer expression.

I've searched with the https://github.com/cplusplus/draft/search?q=%22underlying+type%22 request. This looks like the last place where 'underlying type' is mentioned in a context not related to enumerations or wchar_t/char{16,32}_t.