Repeatablitiy of ERA5? Differences in values between users at the same locations

I am wondering if anyone else has tested repeatability of the ERA5 Land data? Me and some collegues have used the dataset for a couple of papers. But even using data from the same general location we get slightly different values. For example we calculated an aridity index based on dew point, relative humidity, and rainfall at the same area (different GPS points, but same location). For this location I got an aridity value of 1.88, but my collegue got 1.33. Then we had other paired locations (called A, B, C, D) which were 10km apart, so A and C were 10 km apart, B and D were 10km apart, then A and B as well as C and D were about 40km apart. For these locations I got significantly different values for air temp and dew point (but not rainfall) than my colleague. Which then also meant I got different aridity score for these locations. I would not have expected there to be a significant difference since the resolution for the ERA5 data set is about 9km. Any help to understand why this is? How to improve accuracy and repeatability? Thanks