Also, think about the RATE of data collection. If using the numbers above (hypothetically speaking) - they collected 700 million emails per week. Let’s say they actually have 1 million analysts working on these, all day, every day, 7 days a week, 365 per year. That means they’re able to process (assuming each analyst analyses 1 email per minute) 7 million emails per week.
So 700 - 7 = 693 million emails per week accrue (are unprocessed).

Week 1 = 693 million emails
Week 2 = 693 + 693 = 1386 million emails collected and unprocessed by week 2 end.

With 52 weeks per year, what is the absurdly high number of emails collected and UNPROCESSED at the end of each year? (693)*(52) = 36 036 million emails unprocessed at the end of each year. I think that’s approx 36 billion emails. In two years, you get 72 billion unprocessed emails. Unlooked at. Unprocessed. So the numbers become staggering. And these are very conservative estimates. They don’t account for times of high email traffic, they don’t account for the other types of data supposedly collected by the NSA. The claims are absurd. And also - if they’re collecting these staggering amounts of data, when does it become obsolete?
Say they miss an email, in these billions per year, that contains information on a terrorist attack. The next year they find it - is it obsolete? Is the threat still pertinent? How about if the threat is 10 years old? How about if it’s hard to say if the email is a threat or not? Remember, the 9/11 guys talked about “give my regards to the Professor” and “I’m seeing my girlfriend tomorrow” and used all sorts of rudimentary but nonetheless, concealing language. When, and if the NSA analysts - swamped by data - do find one of these emails, how do they determine it’s utility in threat-prevention? Use all emails a week old? a month old? A year? They’d need to analyse them as they came in! Storing them is pointless.

So if we’ve determined that
(a) they can’t analyse all the data
(b) there’s no point storing all the data
(c) to actually assess threats they’d need to monitor emails in real time or at least, within days or weeks of threats being made

then to me, the conclusion is, it would be a waste of time and space, and probably futile, to try and collect everyone’s data, all the time. There’s too much data.