The "disconnected" times are designed to be relatively invisible when looking at the graphs for the remote agent. During these periods where we were unable to talk to the remote agent, we really can't say what caused this inability to talk to the remote agent - and *for sure* we can't say anything about the destination that the remote agent is monitoring.
The best way to get good troubleshooting information about remote agent machine itself is to monitor the remote agent machine directly from PingPlotter. Since you know that the remote agent is reachable most of the time, you can set up another engine type in PingPlotter (call it, say, "Remote Agent Monitor") and then use TCP packets to port 7465. That trace (from your local PingPlotter Pro instance to the remote agent) will show packet loss and timeouts in red.
The grey in the time graph in the remote agent trace basically means "we're not sure about this because we couldn't talk to the agent", so it's appropriate that this not show as a problem period.