Snow depth measurement is a conundrum this time of year, as you all describe here.
My NWS stake is in a shaded area (you can see it to the right of the precip can in the second photo I posted above). I use it as a data point, but not as the data point for depth. I take a handful of manual measurements in that general-but-less-protected area. It gets tricky as I don't want to disturb the snow anywhere near the stake.
I think this is the recommendation from NWS, but the gradient does become significant this time of year. At the end of Feb, when we were just about melted out, I was reporting 3-4" as a depth. As you drove around town, the exposed areas were T-1". My stake was at 5", and most of the other readings were 2-3". Apparently a member of the public contacted GYX to ask about the validity of my readings (like I said, around in non-protected areas it was relatively thin).