Hornet
The product I used to work for (The Capn nav software that all Coast Guard vessels have) got calls about that a lot from the civilian side.
While impractical for the air, we'd ask customers to "capture" the NMEA 0183 data in HyperTerminal, then email it to use. Viewing the data, you could see that the coordinates would be fine, then show random odd readings several hundred miles away, then fine. Our software would endure X amount of these random hits til declaring the GPS input as bad.
If it was bad all around, then we'd speculate about wiring (noise, etc). But to have good data, then bad, then get it back, seemed to just be a bad spot in regards to signal coverage.
I dont know if the AIR GPSs have a means to debug like this, though. Usually these would be posted as known problem areas.