These sorts of patients and calls can run a fuzzy gray line and require good assessment and strong clinical judgement. The problem is that protocols provide a warm fuzzy blanket of removed accountability (hyperbole I know, but it makes the point). Many providers don't have the knowledge and experience to make consistent and reliable clinical judgements, so these get referred to ALS and to ED for evals. All it takes is one bad outcome, preventable or not, to result in cookie-cutter one-size-fits-all policies for these situations.
I personally never approach a patient assuming he or she is drunk. I start from the point of ALOC and start investigating reasons with AEIOTIPS, and in the presence of overwhelming evidence of ETOH consumption, make a judgement that ETOH is the most likely culprit. A former partner of mine instilled this approach in my some time ago, and he's caught a few patients that were having strokes, low BGL, etc because of it.
I don't think the problem is ever deciding a patient is drunk, but rather having it be predetermined.
I'm not sure I actually added to the thread's OP questions, and I hope I don't sound like a blathering idiot, but there you have my approach to "drunks."