No, user stories are not just samples. And they do not target just a bare-minimum product, either.

When collected methodically, user stories represent the language of usage for a given piece of software. It can be as complete as required. A user story set can be as rigorous as an ontology. It is just a different kind of rigour.

Also, you never need a complete set of use cases (just as you never can achieve a complete ontology, either).

I really don’t argue against modeling: it has its place. However, I argue against the modeling bias: assuming it is a universal tool, when it is not. Common modeling practices are not universal, simply because of their hierarchical structure. Many things in language (natural or artificial) are very natural and easy. Expressing/enacting some of those things through modeling is either difficult or impossible.