Because the library gets ported to every new platform anyways (it's what people use so will be the first to be ported to a new architecture or whatever) so the cli gets support for everything new "for free" so can outcompete wget (because it's always _there_ and works the same).
Now, there's also the fact that the curl cli is a fantastic piece of software, but in terms of features I don't think there's that much between curl and wget for simple cli usecases (but I still use curl).
Though wget isn't perfect for mirroring, as it will typically download large amounts of resources redundantly only differing in URL query params when those refer to the same content, such as on typical WeirdPress sites with comment links. Ideal would be an HTTP client with http/3, auth token, and keepalive support etc. based on libcurl that can be customized, such as via "href handlers" triggered by event-driven markup parsers for SGML (though CSS and JS imports need special treatment).
What other common cases take flags on curl but not wget?