Assuming the Garmin is getting proper reception (always a question on anything with an integral aerial, operated in a metal cockpit) it should be accurate to within a few yards.
The land database on a GPS can be hundreds of yards out. Nobody really cares about it too much. Try using an aviation GPS on the road, and you find yourself half a mile off the M25 for example. The only way round that is to run the Ordnance Survey 1:25k or 1:50k chart on a GPS; then you can often tell if you are 5 yards off some footpath...
The airspace shapes are done differently though. They are defined in the relevant national AIPs, as a coordinate list. This coordinate list is lifted and stuffed whole into the GPS database, by Jeppesen (nearly all aviation GPS data is done by Jepp). This may sound like famous last words, but there really is very little room for subtle errors there.
Same with navaid and airport positions; they are lifted from the AIP data.
In reality I think Jepp will use various sources. There is the US DAFIF database, Eurocontrol run a big database of coordinates too which anybody can query. These are machine readable whereas the AIPs are usually delivered printed, which is useless.
Errors do happen but they tend to be major c0ckups, not subtle shifts in airspace.
I'd say a cheap handheld GPS will be more accurate than the ATC radar, too.
There is a different issue with alleged airspace infringements: the controller sees your current track, and if this shows you heading for his airspace, he will get nervous. He may well tell you that you are infringing now. But you could be miles out, just messing about with your track, making him nervous. I happen to know this first hand, from a certain large UK unit. People with autopilots that can track a GPS track or the heading bug can easily make controllers nervous just by switching between HDG and NAV modes; the resulting temporary heading changes (especially from HDG to NAV, if the plane is say half a mile off track) are going to look quite bizzare on a radar screen.