Thermal time constants are "physics". How quickly can we expect soil surface temperatures to change with the adjacent air temperature making a 1-hour excursion below freezing in the pre-dawn hour, when average temperature is 50 or 55F? But in fairness, I suppose most frost is forming on grass, which has little thermal time constant, and also some evaporaton to likely help drive temperature lower.
I guess it's a useless debate. Windshield is icy, weather.com says overnight low is 37F, and current temp is 37F, but there's ice on the windshield. That's sufficient for me to prove they're off by at least 5F. And this happens consistently.
Another data point: For a stretch of a week or two last winter, weather serice was reporting our morning temperature at 0F, plus or minus 2 - 3 degrees. But every morning our vehicles (four vehicles, three brands) were reporting temperatures at -7F to -10F. Again, weather service was reporting 5 - 10F warmer than our vehicles. Temperature sensor on my truck at the time was in driver's side mirror, so roughly 4 feet above the ground.
Conclusion: For some reason, it is always a random 5 - 10F colder here (at least in winter) than what the weather.com service reports on iPhones. I haven't compared often enough to other services, but the few times I've checked, they mostly agree with one-another.