You should not need to get fancy with the interface down alert since the interfaces on a down node should automatically be placed into an "unreachable" status. What I suspect is that your down interface alert trigger is set to fire immediately. That could cause an interface to be polled as down before the parent node is officially set to a down status. If that happens, the dependency would not yet have been applied and you could get the interface down alert.
I would try setting a threshold of 1 or 2 interface polling intervals on the Trigger tab and see if that eliminates the false interface down alerts.
