Considering the number of authors, it might be more appropriate to default to an assumption of competence and basic knowledge of their subject in there somewhere.

I always sort of assumed that since we know stress is a major factor in health, that moderate drinking probably has an overall beneficial effect due to the stress reduction, even if the physical effect of the alcohol itself is slightly negative. Like maybe you get a -2 modifier for consuming alcohol and a +5 modifier for stress reduction and so you end up ahead. Same goes for the likely added benefit to social interactions, unless you're drinking alone in the basement.

Then again, I'm pretty far from a scientist so I'm just spit-balling here.

Right? Two obvious causal links would be stronger and more numerous social ties lead to both increased social drinking and decreased risk of dementia.