Doesn't this mean that their selection sample effectively excluded any babies that had a noticeable physical reaction (other than having cocaine in their system) to the effect of their mothers' cocaine use?
It is a logical thing to study, and most likely is being studied, by someone else. After all, this study wasn't looking at premature babies. That doesn't prematurity with crack isn't a problem, only a different problem.
I imagine somebody who is actually in this field would not have much trouble designing experiments to study premature "crack babies". Premature babies are something they chose not to study, not something that cannot be studied.
Excluding premature babies of course limits the study's relevance, though the results may be remarkable even if they only apply to full-term babies.
A study of only premies would be interesting, but might yield the same result as this study.
The rate of premature birth among all pregnant women is around 4.3% (that includes smokers and drug users so the rate amongst healthy non-smokers is even lower than that). Amongst crack cocaine users the number is between 17–27%. So the prematurity issue is pretty significant. You might even say dominant.