It's not settled because there are still delusional people clinging to the notion of human capital development, despite the almost complete dearth of evidence.

My recollection from the enormous volume of papers we covered in grad school on this topic is that basically every different approach has ended up being more consistent with signaling or filtering than with human capital development. You end up needing to lean on some really odd and implausible assumptions to maintain the human capital story (like all learning happens right before graduation).

What I will say for the human capital hangers-on is that their treatment variables have much more measurement error (we only ever see imperfect proxies for amount of education received), so we would expect the estimates to be attenuated and harder to register.