When I tried making sense of this topic in the context of the controversies over IQ heritability, the best reference I found was this old paper:

Brian Mackenzie, Fallacious use of regression effects in the I.Q. controversy, Australian Psychologist 15(3):369-384, 1980

Unfortunately, the paper failed to achieve any significant impact, probably because it was published in a low-key journal long before Google, and it's now languishing in complete obscurity. I considered contacting the author to ask if it could be put for open access online -- it would be definitely worth it -- but I was unable to find any contact information; it seems like he retired long ago.

There is also another paper with a pretty good exposition of this problem, which seems to be a minor classic, and is still cited occasionally:

Lita Furby, Interpreting regression toward the mean in developmental research, Developmental Psychology, 8(2):172-179, 1973