The sensor I used to have was also on a north facing fence. I noticed that it was incredibly sensitive to cloud cover on a cold night and the temperature generally fluctuated much more quickly than properly screened sensors. It gave readings that were too low. This is supported by posts in this thread, which coincide with my experience.
https://forum.netweather.tv/topic/63499-do-i-need-a-stevenson-screen/
As you can see, non-screened sensors in relatively open locations can lead to temps that are up to 2-3°C too low on calm, cold nights.
Most PWS were reporting higher than you got, and if one or two weren't, then it may be that they aren't screened. Certainly, the only official station that was close was the airport, which is very exposed.
The main point is that without a screen, readings (on such nights especially) are unreliable.