I have a G-Tech Pro (revised version of the first model). Anyways, a couple things come to mind:
1) Not very accurate for fwd or awd cars (on my unit anyways). Reason being I suspect is that for the G-Tech to start recording, the accelerometer has to start detecting acceleration. It does this by the accelerometer tilting inside the unit when G forces act on it. Now with a rwd car you will get quite a bit of weight transfer and thus pitch. So on my MKII, the G-Tech seems to almost kick in as soon as the hammer goes down (launching easier will give a delay though). On the other hand on my fwd Celica, or my freinds awd Talon, there seems to be a delay of several tenths (even upto 1 sec sometimes), as they don't seem to pitch as much due to their drivetrain layout. This discrepancy throws off the units accuracy, and accounts for the quicker times as some have related above.
2) The initial G-Tech also only has one internal accelerometer, therefore you have to "zero" the unit first before you can start using it. If this is not done properly, it will throw off the results. Although I have gotten good at doing this quickly, sometimes it seems like no amount of tilting the unit will "zero" the damn thing (probably has to due with the units friction hinge loosening or item #5 below). With multiple accelerometers, I would assume the newer G-Tech Competition model is alot more accurate (don't need to zero it), and might have that initial "time delay" thing solved for fwd and awd cars on the drag strip.
3) Accurate weight input is a must.
4) For the reasons mentioned above, the hp values cannot be deemed as gospel either. I tried different launching techniques with my MKII, and got different hp values each time. I would also get different hp values depending on wether I was shifting normally, or speed shifting. The engines hp doesn't change, but depending on how you drive you will get different hp values from the G-Tech (old one anyways, not sure about the newer one). Therefore I would only use the hp function as a method of doing "before and after" tests when installing performance components, and do it from a roll in one of the higher gears (so factors like clutch slip, and shifting "jolt" don't throw the unit off). This won't give you an accurate whp value mind you, but should give you a somewhat accurate hp "gain" or "loss".
5) I don't like the big suction cup used to hold the unit to the windshield. I'm sure the suction cup/s "flex" when the car is bouncing around while accelerating or cornering, which would also throw the results off.
6) Having said all this, it's still kind of a fun toy to have (as long as you don't get carried away with what the results say), and am curious to see how well the new unit works. My 2 cents worth.