There are at least two limits
- visual magnitude that could be measured
- positional accuracy that could be measured
The positional accuracy eventually was on the order of 0.001 arc sec (or 2.78x10^(-7) degree) so from that you could naively work out that the distance that could be determined. (Measure at time A, subtract measure at time B occurring 6 months later, best measurement difference able to be measured is 0.001 arc sec)
Naively: If I recall my simple trig correctly, for the purposes of first order feel gives: tan(2.78x10(-7)) = 1(AU)/h where h is the distance of the object in kilometres and 1(AU) is the earth sun distance which is your maximum baseline. So with everything working in your favour you could get out to about 2.061x10^(8) AU or 3260 light year.
In reality you have plenty of other factors involved including how the data is taken and you also are measuring against a moving background not an absolute one. Further the satellite was after accurate positional measurements of the stars first. So all this means is that the actual figure is out to around 1600 light years.
Whether or not all stars within that volume were bright enough to be accurately measured and were part of the observing programme is the completeness question.
There were also other confounding factors such as whether the stars are double or multiple.
There is a satellite programme due to go into orbit (Gaia:
http://en.wikipedia.org/wiki/Gaia_mission) which will seek to accurately measure a huge number of stars and then solve the very big inverse problem i.e. solve all positions at the same time due to the various inter-relationships. Good fun debugging that code. The target is to get out to greater than 10,000 light years.