When doctors tell you normal body temperature is 98 degrees they are dumbing it down for you, instead of relying on bad sampling from the past.
Not that medical science is that great either.
Suppose a station uses a mercury thermometer for the first 80 years then switches to a thermocouple, with an overlap of 10 years were both thermometers are run side-by-side. Suppose also that it shows that the old thermometer consistently reads 0.25° warmer than the digital one, and that calibration tests show that the new thermometer is accurate.
Do all of the old data points get thrown out of the window? What's wrong with adjusting the old temperatures by 0.25°?
There's whole disciplines in engineering that deal with getting good data from imperfect sensors. Have you heard of sensor fusion? Whole field that uses multiple sensors to mutually correct error. Have you ever wondered why digital barometers come with a thermometer? No, it's not because it's convenient, it's because a raw digital barometers reading fluctuates with temperature, so the temperature sensor is used to correct the raw output.
https://arstechnica.com/science/2016/01/thorough-not-thoroug...
Every data from every experiment in every field has to be adjusted, raw data is meaningless