One current limitation of NPM relates to the ability to monitor thin wireless access points (at least when they're connected to a Cisco wireless controller). Currently when those APs go offline, they go into an "Unknown" status for one polling cycle and then disappear entirely from the platform. As I understand it, however, it is possible to see disconnected access points via "show" commands on the controller even after this happens. This essentially means that successful alerting is dependent on a race condition where the alert must be evaluated and fire within one polling cycle of the AP being detected as offline.
What I would like to see in future versions is that once a thin AP is detected, that it is stored persistently in the database. If it goes offline, it should continue to be listed in an offline status until it either returns to service or is manually removed by someone with node management right.