In practice, "black swan" is supposed to mean effects which we underestimate or fail to predict at all. The definition requires not just rarity but an event which hasn't been seen before, and is only only understood in retrospect.

You can try to leave a general hedge for unknowns, like never giving anything >99% long term confidence. But actually pricing black swan problems accurately should be impossible, pretty much by definition.

It just so happens that people, when making predictions, are not very good at accounting for ("pricing on") those kinds of events.

Ha ha. No they (the people who make real deals/trades/decisions) are just focused on their bonus. Having said that they also care a bit about their bonus. And of course their bonus is important too. Have I missed anything?

I think you're mostly comprehensive there. You missed out two things though: they care about their bonus, and secondly, how that bonus compares to their peers (which could be summarised as "their bonus").

After which they blame the "bad model" and immediately pass on to constructing newer, presumably better models. The fact that maybe, just maybe, trying to model such a thing as a modern-day economic system is futile never crossed these people's minds.

You are being unfairly downmoderated. It is of course not a given that modeling economic systems is impossible, but there is considerable debate as to e.g. what degree prices are random/unpredictable. Everyone agrees the problem is extremely difficult, and it seems somewhat unlikely that there would ever be such thing as a universally applicable financial model. So the remaining question is, of course, to what degree are these models useful? I do not believe there is sufficient evidence to reject the idea that the models are of no utility, especially in the context of failing to price in risks. The contrary position is equally viable, and I would hope anyone with what they considered to be a knock-down argument would do more arguing than knocking down.