What I think I'd say (though I'm not entirely sure it's right) is: (1) as you say, we can't know in advance how freely the person will choose, (2) more people will make the much-easier choice (that's almost what "much easier" means), (3) in some sense the right way to think about how much freedom that choice itself introduces to the world is to look at the average -- it might help to imagine millions of people making similar choices, so that most of them make the easier decision, etc. -- and (4) the average works out so that a more-pressured decision introduces less freedom to the world.

I'd say about the smoker either "coo, well done, that was a really difficult choice" or "ah well, it was almost inevitable; try again". (It wouldn't be "it was inevitable, he wasn't free" unless there was really no possibility of choosing to resist.)

On reflection, I think I dislike some terminology I've used here. The quantity that's different in the two cases isn't exactly "freedom", it's something more like "decision". Freedom is (at least according to my analogy) expected quantity of decision.

I don't think the free-ness of a decision can really be affected by whether or not there happen to be lots of other people making very similar decisions, so I'm happy saying "this is how it is in the many-instances case, so that's also how it is in the single-instance case". And since I'm strongly inclined towards the "many worlds" view of QM, maybe there are in effect *always* many instances, with many of them going each possible way...