Source. Open-Meteo's historical archive: ERA5 reanalysis from 1940 on, with ERA5-Land detail where available, downscaled to the spot's elevation. It is a model of the atmosphere constrained by every observation available at the time, not a thermometer at Hyak. Two things follow. Records run milder than a station would log, because a grid cell averages the valley floor with the ridges, so a single cold hollow or a sunny slope never shows up. And anything before about 1979, when satellites entered the assimilation, rests on far fewer observations, especially in the mountains.
Normals use the most recent 30 complete years, the same span the NWS uses for its normals. The climate has warmed enough that averaging from 1940 would describe a colder place than Hyak is now. To smooth out single-date noise, each day also pools the 7 days either side of it, so a normal is drawn from roughly 450 days. Records use every year on file but only the exact calendar date.
"Most ordinary" is the actual year whose weather on that date sat closest to the normal high, low and precipitation together.