Sure, having some variation in circadian rhythms among the members of a group makes it easier to have guards at most times of day and night. But such variation exists independently of age: There are "owls" and "larks" at all stages of life. I find it interesting that, on top of that, teenagers specifically and consistently have their individual rhythms shifted towards the later hours.

That's a really interesting point I hadn't considered. And thanks to evolution's pseudorandom algorithm, it could be that the benefit was obtained in conjunction with another benefit (nighttime socializing), that one came about after the other, or that one exists and one doesn't.

Can the purpose of this mechanism not be determined by observing hunter gatherers?

I don't think there's that many of those left, at least isolated and big enough to serve as an actual testing group.