Yes, Garmin says that in their one sentence blurb. But, when you dig into it, it's a lot more complicated. You have to understand the nature of sensor calibration errors, the various contributions to GPS error, and how all those errors accumulate over multiple measurements.

You're missing the point. The most accurate way to get is good number for the roll out is with GPS. That's why the sensor cannot be better than GPS at distances greater than the distance used to determine the rollout.