Here's the thing: I bet if you were to take the children who performed well at complex abstract mathematical problems and placed them in a job that required complex applied mathematical skill, the ones who eventually performed best would be the same. And vice-versa for the children who excelled at complex applied mathematics. So mathematical skill might not be immediately transferable, but would be a useful indicator of innate ability.

This relates to the (somewhat controversial) economic theory that much of higher education is not about training or creating transferable skills, and instead is about "signaling". Essentially, we place children in simulated quasi-work settings and use it to determine who has the most of certain economically valuable abilities. By performing well at abstract mathematics, a child signals that they have the requisite mental ability to tackle similarly complex "real world" problems. But also, they signal that they are capable of obeying authority, working with diligence, delaying gratification, etc.