What puzzles me is the reasoning that users choose #2 because of that. Clearly the choice should be #1, or not?

Well #1 has to exist (which requires creativity and initiative on the part of the posters, so you can't assume infinite quantities at all times) and be identifiable. Certainly some of that does. I vote for it. It also in many cases (not all) has a finite capacity: If a post is already fully rewarded to the extent of its value by the time you considering voting on it, you can't vote for it, even though you might entirely support it getting the rewards it is already voted to receive. (An exception would be posts which are raising funds for something with a large or essentially unlimited budget.) This gets into the marginal vs. total question I alluded to earlier.

"Not" (by which I guess you mean not voting) is only a good choice if you think that your vote for none-of-the-above is causing more marginal harm by reducing rewards from #1 than marginal benefit by reducing rewards from #3. This is obviously a judgment call. If you think everything is working great then clearly you would have no reason to vote for #2. But IMO that's rather absurd given Steem's lack of success on almost every front: A system which effectively allocates budget to where it can do the most good and is most effective in drawing value into Steem should be akin to a superpower, particularly when few competitors have this. We're clearly not doing that.

In an ideal world: #3 wouldn't exist (or would be fully downvoted) and #2 would only be a sensible vote if everything in #1 were already fully rewarded. But we don't live in an ideal world, at least this blockchain and its rewarding system certainly does not. #3 routinely does not get downvoted because of too much drama, too much unrewarded effort, etc.

Also, most of these judgments are subjective in terms of what constitutes value and such, and that's okay. I am less concerned about people disagreeing on what or where is the positive value in good faith than people gaming the system to extract value from it or people not downvoting when the latter is going on (which then ensures it will, because there are always people who will take advantage when they can). That's a systemic problem here. Since we don't have a solution, none-of-the-above (i.e. burnpost) goes up in value from where it would otherwise be if we had more effective downvoting.

Likewise for curation initiatives. Sure there might be some excellent curation initiatives but I don't really see a good way to figure out what is good and what is not (is there anyone curating the curators?), and for the most part I am suspicious of them given past experiences and some of what I still see today. Finally, most of them don't downvote or don't downvote effectively, which is yet another reason for me to not support them and retain control over my own vote power instead. Still, if someone is able to convince me that this or that curation initiative is doing a great job recognizing value, then I'm open to being sold on delegating to it. That doesn't happen by default. When I support burnpost, I see that as an case of "do no harm" relative to delegating to a curation initiative that I don't know is helping or hurting.

Finally, I'm also comfortable that voters can decide how much of the pool should go to none-of-the-above. If burnpost gets no rewards for a long time I'll probably stop posting it. If it gets small rewards or large rewards, IMO that is just people making a budgetary decision.