The point of using individual rings -- something made possible as the equipment became more sensitive, thus able to test much smaller samples -- and the annual variance of the RC accumulation is calculated. The sequence of values is then matched up, "wiggle-matching", so that there's a continuous sequence of data even though no one tree covers the whole thing.
So they eventually match it up with a tree they can actually date? Seems like a lot of error might be able to creep in with this methodology. “Wiggle matching” seems like an art form.