Drug metabolism is often better modeled by two (or sometimes more) different half-lifes for different stages of metabolism. So typically plasma concentration will diminish quickly but be followed by a long tail. In the case of caffeine I think this is due to enzymatic saturation.

It seems that this second half-life must be the difference for people who have more trouble sleeping long after caffeine intake. I wonder if this implies that their bodies have different maximum attainable amounts for the proper enzyme. I'm curious if anyone knows more detail about how the interplay between enzymatic saturation and genetics / gene expression works here.