Were these 2 APs ever a part of another swarm? If they were, then this sounds like a known issue, and the only workaround is to factory reset any AP that has moved from one swarm to another.
This has been fixed in the 3.0 version of Instant.
In case you're interested in details: each swarm sends a unique identifier called a guid to AirWave so we know what swarm is contacting us. But in pre-3.0 versions of instant, APs that move from one swarm to another maintained their old guid. When 2 VCs contact AirWave, each reporting a different set of APs, AirWave will alternate which APs are up depending on which VC was the most recent to contact us.