It took two years, but I managed to get a job at Stack Exchange specifically to investigate this for myself, that's just how dedicated I really am.
The answer is: (drum roll) ...
It's recorded, just like any other flag, but the data isn't really analyzed. If you think that an auto-generated flag isn't helpful enough, isn't helpful at all or should be tweaked so that it can be more helpful, raise a feature-request here to discuss it.
These are auto generated based on specific criteria that we've learned is probably worth investigating when it surfaces. Remember, these flags aren't the Community user telling you to go do something, it's just suggesting that you take a peek to see if something needs to be done. Declining these flags has no effect on the flags it raises, or how often it does so.
What is helpful is when users that have the privileges to do so dispute these flags when there's nothing worth investigating. This helps the moderators spend less time looking at the post wondering what Community is squawking about after seeing that several trusted users found nothing.
Still, even when there's no metaphorical fire, these flags are accurate, as they're just the artifacts of pattern matching at work.
And, even though I've fulfilled my long term quest of getting hired so I could see this for myself, I think I'll stick around.