Sometimes a down-time event is recorded due to misconfiguration internally to the url that uptime robot checks, however the actual service remains accessible. This is especially true for internal processes whose status is exposed via another mechanism.
In this case the actual service is still “up” but uptime robot doesn’t give us a way to either remove or disclude a specific event from our statistics.
We’re trying to improve end-user transparency, however when something like this happens it “dings” our statistics even though no actual down-time occurred.
There are even cases where the ISP that serves us has their own issues, which are beyond our control. Again, the service is up and still accessible, but the connection between us and uptime robot is down.
Adding a way to flag or remove a specific event so it does not ding statistics would give a better picture to end-users.