90-95% of NAA flags are reasonably accurate. Compared to VLQ at 60-70%, that's really quite good. Because of this, they're not normally that much of a burden on the moderator team, at least in contrast to VLQ and especially "other" flags.
Inline sidebar: Not An Answer flags during the past 30 days on Stack Overflow
NAA Helpful On Now-Deleted Post Helpful AND on Now-Deleted Post
----- ------- ------------------- -------------------------------
57203 52733 52860 52281
Either way you slice it, that's over 90% accurate.
That's not to say this couldn't be improved, however. A big problem right now is simply that they're exposed to 10K users, but 10K users can't vote to delete them! They can dispute them or pile on more NAA flags, which can help the moderators process them more quickly in some instances but can also just inflate the flag count (this was a particularly bad problem during Winterbash and the recent moderator election).
We could do something similar to what we did with VLQ flags (direct them into a queue where a large number of low-rep users are allowed to delete them) - but this is kind of a waste when only 1 out of 10 won't be deleted.
A better option might simply be to put them in front of 20K users and give them the ability to delete OR decline, with only controversial and unhandled going before the mod-team.