To be fair: it might be they have never done this before.

At my previous company, we had a technical assessment - this was about ten years ago now. It boiled down to: read XML, do some math / business logic, and build a REST API to do so.

Interestingly, ten years ago, at least half the applicants said they found it interesting because they had never worked with REST or JSON before. A lot were Java developers, so the XML part wasn't a problem, and they would often add some SQL database as a bonus.

But 5-10 years later, as development switched to (Node)JS and web, it became the inverse and people said they had never done anything with XML before.

And far more likely: done before, but never on anything even remotely close to a blank slate. You can spend years doing X, be very good at doing X, but only ever adapt some pre-existing precedent implementation of doing X to a new use case, or to a new underlying library, but never any green-fielding. That "implement X in a vacuum" test will rate many experienced people lower than some who have never ventured beyond textbook examples. It's not impossible that your real tasks have so much green field work in them that those experienced brown-fielders might actually be bad matches, but I suspect that those situations are much less common than the tests that select for green-fielders.