I fired up my scatter plotter (a work in progress) and collected 24 hours of fix data from my GM-2. I connected it through two six foot USB extenders and placed it atop a ladder on my deck attempting to get it high and out of a multi-path environment.
WAAS was turned off. The GM-2 was operating a one fix per second sending at the default 115200 baud.
The position in the lower left of the plot window is the average of the fix locations and it agrees to within a few feet of the satellite image position of Google Earth. That's the best I can say about accuracy within the methodology of the experiment.
A CEP (Circular Error Probable) circle contains fifty percent of the fixes. It's fairly accurately calculated from the radius error histogram. For this run, that circle was only eight feet in diameter.
Oddly, the receiver reports position with six digits of minute precision i.e. 41 deg 03.589058 min or, less than an inch. Not that it matters but reporting that precision raises those arguments concerning the relationship between precision and accuracy. But, it renders the scatter plot as a 'Ball of Yarn'.
Magnetic Variation is not reported as indicated by the two empty fields in the NMEA GPRMC sentence: