A nagging thought on NSA surveillance, not terrorism related.

[alc]

2. don't want to expose their methods

Every time they use their surveillance they show that they have capabilities. If they use it in courts, the defendant has the right to know how they got it so their expert witness can dispute the validity (someone else may have used the computer/phone). Even without courts, they tip off terrorists when they use it. If a drone strike hits a secret meeting that was only discussed on a couple of channels (phone, email, etc) the terrorists know that channel was compromised. That's another reason we only want to take out the top terrorists and not anyone we can.

The other issue is manpower. The computer analysis most likely spits out 1000s of potential threats a day (if not much more) just for potential terrorism. That's a wild guess on my part, but I do have experience in big data analytics and visualization and have see reported numbers of terrorist attacks have been averted and assume the goal is to stop ALL attacks (number of false matches you need to manually rule out depends on how many missed matches you are willing to let through). When a person looks at the computer matches, some are obviously false hits but others may take hours or days to rule them out. If they were looking for potential criminals, they'd likely have millions of people per day to look into in greater detail. They cannot pass millions of "leads" off to the FBI every day.