Several things can cause that. Remember that SNMP rides on UDP which is an unreliable protocol. Most NMS tools (including PM2.0) have retries and timeouts to mitigate this. So, this could be as simple as packet loss that the timeout/retry couldn't recover from. Does this happen across all trends at the same time or is it limited to this one device? If limited to this one device, it could be that the device was too busy to respond. If the same gap appears across all devices, it would appear to be a problem with anything they all have in common: common poller, common link to the network where the poller resides, data aggregator, data repository, CAPC.