Sounds like a salespersons's smooth talking about GPS' capabilities in general versus the "fine print" on the actual product itself. For civillian use, isn't GPS accuracy in the 3m/10 ft range? Which would jive with the guy's assertion of what he was told.
The ~10 m range is what I originally thought too until I looked at the Cospas-Sarsat spec's for the transmission from the beacon. I haven't taken the time to completly read that part of the specs, but If I understand correctly, the location information transmitted back by the beacon is limited to 4 s of lat/long in the best case - it could be worse like 2, 4, or 15 minutes of lat/long.
4 seconds of longitude works out to about 123 m for equatorial latitudes and around 95 m for mid-North American latitudes. 4 s of latitude is about 123 m regardless. So figure a something a little larger than a 100 m x 123 m area as a best case fix.
If it's 4 minutes then the area grows to 5700 m x 7400 m for NA latitudes.
That's where the 121.5 MHz beacon starts to come in handy for RDF location of the beacon by local SAR folks.