I will attempt to put clear the problems with the 'It can work' argument in the form of defining what the best case scenario is, and what must be already in place to acheive it.
I concur with your points.
Torture is only a (potentially) useful recourse in situations that are so contrived they will probably never occur. The barriers to torture drop under certain conditions (say, the size of the area that the security apparatus needs to cover in order to determined the truth of each of the victim's confessions), but at no time do they drop even remotely near to the point that we should be asking ourselves whether we should actually drop prohibitions on torture.
Doesn't need to be a liar. Just needs to be a statistical outlier. That's why singular cases aren't discussed, why anecdotal evidence isn't actual evidence.
It is a statistical outlier. I disagree that we shouldn't discuss them. If you were diagnosed with a terminal illness and you could try a treatment that had a 1.0% chance of success, but had exhausted all other options near to the end, what would you do? I agree that it's very hard to think of a situation where that might be the case. But is it impossible? How about the German child case?
Also, what would you do if you were Bill Cowan or a member of his unit? Would you let the North Vietnamese prisoner go? Knowing that your men know the location of similar traps? How about when somebody reminds you that the North Vietnamese torture Americans even when they know the Americans do not do the same (routinely)?
Personally, I have a hard time faulting Cowan.