Just saw that, thanks for the link. I stand corrected, I hadn't personally seen them stay very still in a hover before, maybe it was operator inexperience.
If the quad has only an accel and gyro sensor onboard then yes, the stability will be based primarily on the skill of the pilot and how well the machine is tuned.
With a GPS and barometer thrown in, position hold should be pretty close to a still hover unless there is significant wind or unresolved problems with the power train
The copter doesn't need to know where it is, just whether it is moving and if so, how far and in what direction. Because the error in GPS measurements doesn't vary much in small time frames, that can be done more accurately than absolute position measurements. See http://en.m.wikipedia.org/wiki/Differential_GPS.
A "good" GPS system (the DJI version is ~$100 by itself I think) should be able to get you down to just a few inches of precision, but yes, the market is flooded with cheap (~$20) units that aren't nearly as useful for position hold (hovering completely still) but can still give you a good lat/long and do well with navigation, distance measuring or acting as a reference sensor for a automatic direction tracking antenna. Cell phones are similarly hit or miss-- some have nicer GPS modules than others, but ultimately the user is unlikely to notice this since smartphone style road navigation only requires 10 meters or so of accuracy and a vague single position can be greatly refined in that context by knowing the road/map layout as well as the direction and speed of travel.
With a GPS and barometer thrown in, position hold should be pretty close to a still hover unless there is significant wind or unresolved problems with the power train