While I wish I would have jumped into this thread earlier, I don't have a big interest in rooting through page after page to catch up on how this discussion began or what the two sides of the discussion are arguing. That said, I would like to respond to the bolded portion above.
I would imagine incarceration for any reason is also highly correlated with diminished lifetime achievement. Given that the US incarcerates people for weed use, I'm not sure this is a surprising correlation. One positive test would have knocked me off of my career pathway. Would a positive test have made me (on an isolated individual basis) any less capable of maximizing my potential lifetime achievement? No, probably not, but given the fact that my punishment would have been for my career path to be derailed it's not surprising that such a correlation is out there. Consequently, the correlation seems like it could be based more on extrinsic societal, legal, and even psychological (in the form of hiring practices) factors of marijuana use than it is any sort of intrinsic aspect of marijuana itself.
I think the case with Randy is that he has a whole list of issues in addition to a dependence on marijuana. Unfortunately for him, while being marijuana-dependent is likely the least of his problems, it just happens to be the one that he can be punished for. He could have a list of behavioral, emotional, and/or psychological health issues that undoubtedly contribute more to how far he gets in life and never miss as many games as he's already missed for testing positive.
In any event, I don't think Randy is a good example for any of the outcomes that could be cited in literature because he either:
A) Has behavioral, emotional, or psychological issues that make any sort of comparison questionable, or
B) Doesn't have any underlying conditions, in which case he has a level of dependence that far exceeds the norm. With over 40% of the US reporting use at some point in their life, Gregory is undoubtedly an outlier within the outliers. Apply such a level of dependence to almost anything and you'll be able to draw some negative correlations.