> Nothing about the system would have been compromised by publicly announcing they had detected an anomaly the day of the event vs 4 days later.

Have you worked on classified detection systems? Actions and conclusions don't always appear to follow logic when your priors are wrong.