The software complains about this whether it's relevant to a given application or not, as it can't know. If you use a combinationally-generated signal to clock a flipflop, you should expect to see that complaint. It's the designer's responsibility to understand the consequence of such a ploy. It's clearly asynchronous.

In a case wherein you're building a large synchronous block it's unwise to do such a thing. If it's just a push-on/push-off control for a LED, it's not so serious.

Real circuit design is seldom entirely synchronous, that is, there are often multiple timebases involved. The crux, in such cases, is in making everything minimally risky. These warnings help direct one's attention to such things, even when it's really not warranted.

RE