My recommendation would be to first determine how long after an outage of a projector is an acceptable time-frame before considering an outage a new outage versus still part of the last outage.
Depending on how like that window is, I would say to follow @b0fh suggestion and redesign the check. If the window is short (several minutes) simply tell the check on an OK result to sleep for X number of minutes and then rerun the check; if it passes the second time then send the exit code 0. However, if the window is longer then several minutes, I would say a better option is to redesign the check essentially with status caching (so that you can compare an outage/device up against the cache). In order for this method to be more effective, you may need to run the script/check as a scheduled job on the Nagios host and have it send passive check results to Nagios.