You argue that is incorrect? What's illogical about that statement? ...and what it looks for yet you make the pronouncement there's 99+% noise.
In response to your implication that people should cow to 'the list' for fear of winding up under special observation by simple mention of particular words. If they are consuming all content from all of the major websites, application of those 'watch words' against the whole of it with no other filter would produce a useless mass of false positives, because all of those words are very common in innocuous news and commentary, etc. But I'm just a software architect, so what do I know about figuring SWAGs and feasibility from requirements.
They are amassing incredible amounts of data and increasing their capabilities all the time. That is certain.
Nobody is arguing that, or that it's an unnecessary overreach and should be stopped. The point of contention is simply the need for rational analysis and response, rather than FUD.
Perhaps they're giving you disinformation along with crumbs of truth or near truth about the programs?
Nobody is arguing that either. But as with anything, you do yourself a disservice by giving in to paranoia and hyperbole, rather than applying critical thought. The simple fact is that what has been reported in official statements as to how it is used (including that very article about the watch words) makes a lot more logical sense to anyone who knows how databases actually work than do the exaggerated headlines. Sure, it may (and probably does) do other stuff, but the core function and architecture of the system still has to follow the same rules as any other data storage and analysis system does.
If they had made some spectacular breakthrough in data analysis such as people attribute to them, I have a very hard time believing they would be able to keep it to themselves for years. Because such a thing would be worth a very, very large amount of money in the private sector.
If you, one of your friends or associates goes haywire, they'll know very quickly who to approach and question about you
They can already do that by looking at public information on Facebook alone, for most people; they have already gladly waived any sort of privacy on that sort of thing. If anyone using Facebook weren't already assuming that were the case, then they're quite simply fools.
What is clear is that it doesn't have a very high predictive ability.
Which is an argument for it being not omniscient as you fear. And people who genuinely intend to do harm on a large scale know that well enough to discuss it offline (or not at all, like the leaker).
There's also the case of the Tsarnev brothers, on whom they were even given warnings by foreign intelligence... and still failed to stop.
TL;DR: We agree it needs to stop, but at least apply some critical thought instead of believing (and more importantly, spreading) all the hyperbole you read. If you can't assess and respond rationally, you can't effectively contribute to stopping it.