Yes, this is true. Pull up at a junction and your sat nav will often still read 20 metres/yards to the junction so as stated above they have downgraded accuracy in them.
Also the Ordnance Survey now use satnav to plot their maps, they plug a lazer reader into a trig point and it scans millions of points in no time at all. Ordnance Survey requires tolerances of 1 metre at 1/1250 scale, so they obviously have a much more accurate system too.
(Although Ordnance Survey do indeed use Lasers extensively, the Laser is used only for rangefinding ie. distance measuring and does not as such use GPS technology, although an accurate GPS receiver will tell them where it is.)
With regard to GPS/Sat Nav think there is a confusion here between the accuracy of the GPS transmission and the quality of the receiver.
Selective Availability was turned off at 12:04AM 2 May 2000, US Eastern Daylight Time - that is fact. This allows a suitable GPSr to resolve locations to easily within 2 to 3 metres. However this always assumes that you get a good satellite lock.
Surveying GPS uses a further fixed ground station for error correction and of course the measuring point does not move at all - this will actually allow accuracies of 1-2 centimetres, but is not technology that lends itself to vehicle navigation, where a car is in motion. They also use a technique called Carrier Phase Tracking where they use special receivers to measure the wavelength and frequency of the L1 carrier signal at a fixed point, and with a 1% accuracy of the wave, this can give an accuracy of 2 millimetres.
Companies such as Racelogic use a ground station system for precise data logging for racing cars with high sampling rates for accurate tracking and data recovery on twisty racetracks.
Military GPS uses both the L1 and L2 channels for increased accuracy - civilian consumer units are denied the use of the L2 channel. By using two channels, the effect of changes in the ionisphere can be compensated giving greater accuracy. It would be fair to assume that the sampling rates of their receivers are also extremely high and they can use Doppler shifts for more accurate velocity measurement. So yes it is true to say that military systems are more accurate - but this is entirely due to the technology of their receivers rather than the GPS signal.
From that you can see that the GPS satellite signals received on the earth's surface are incredibly accurate and precise and are now NOT degraded for civilian use, although we are denied access to the L2 channel. As you read this there are both L1 and L2 signals passing around you!
When it comes to receivers, it is quite possible to go out and purchase a GPSr with the level of accuracy stated above. Error-checking systems which overlay the GPS networks and correct for ionispheric disturbances, atmospheric changes and intereference, such as WAAS in the USA and EGNOS in Europe, enable moving vehicle accuracies within the range of 1-3 metres. However, sat-nav manufacturers are a bit coy in describing whether these technologies are used in their consumer units.
Without the benefit of EGNOS overlays, 2D approximations and no ground station checking, even the cheapest of the consumer Sat-nav systems will give a reasonable approximation of your location within 5-15 metres, say a road width, which is more than adequate for their purpose. They also use slower sampling rates, typically 1 Hz (Once per second) and smoothing algorithms - all of which lead to small errors like stopping short of junctions and appearing to continue moving with the vehicle at rest. However, this same technology with some sat-nav fairy dust, is also what allows you to drive through a tunnel or under a bridge - so don't knock it too hard.
With regard to the effect of hills - if you were receiving a signal from just one satellite positioned directly overhead, then your ground speed could be seen to fluctuate as you went up and down steep hills. Anyone who used a Sat Nav in their car will know that this does not happen - at a constant road speed of 40 mph your GPS speed will not vary between 35 and 50 mph as you go up and down hills for example.
In the real world, you are tracking between 4 and 12 satellites at varying heights in the sky and the GPS world is actually in 3D. Add to that the Topographical information collated by the map makers and you can be assured that a GPS receiver knows where you are and what you are doing. Again, the buffering and sampling rate of the receiver is also significant in this, smoothing any small fluctuations. (That is also why standard Sat-Nav GPS is not so useful for acceleration tests because your receiver has a built in lag and averages small variations). It is also fair to say that Sat Navs use PFM (Pure F****ng Magic) to make them work - it is an incredible technology.
Real world velocity accuracy of most consumer sat navs - worst case in all road conditions and terrain with a minimum of 4 satellite locks, is around 3-5%. In other words, between 2 and 3 times the accuracy of your speedometer.
Be aware that there is a crisis at the moment in replacing older satellites which may start to die before replacements are launched (can't see the US Military letting that happen) and that as the entire system is under the direction of aforesaid US Military they can turn off selective areas of the world at will, to prevent terrorists using the systems, or turn it off altogether if they wished. Next generation satellites have specifically had the Selective Availability feature removed, as the civilian benefits of GPS have been recognised and more users such as airlines have become dependent on the technology. USA and Europe are testing and certificating GPS controlled automatic landings.
So if your sat-nav leaves you short of the road junction - its your receiver NOT the system.
I now have a headache so if you will excuse me I am going to lie down in a darkened room