I'm not convinced this is even true:
1) The effectiveness of dragnet surveillance is vastly overstated. There are far too many haystacks for the number of needles.
2) Even if it was effective in a vacuum, there is no evidence that alternative methods of fighting terrorism would not be at least as effective. Even if you stipulate that the amount of resources being spent on anti-terrorism is reasonable (which is by no means a safe assumption), using those resources for e.g. human intelligence may in fact be more effective in preventing attacks.
3) Dangerous automated systems are dangerous. Never mind federal agents stalking their ex, what happens when the terrorists you're trying to thwart break into your dragnet surveillance apparatus, or make use of the same vulnerabilities you've built into the infrastructure, and use it to determine the schedules and defenses of targets in order to plan an attack?
They are creating a security vulnerability in the hope that it can be used against the bad guys while ignoring the possibility that the bad guys can use it against you. This is not such a good idea when you have a bunch of obvious soft targets all over the place and the bad guys don't.
Which is before your quite valid argument that government officials may be among the set of people capable of misusing such power.