As discussed here (https://ec.europa.eu/jrc/en/PVGIS/tools/tmy), PVGIS is generated from a combination of satellite-based solar radiation database and ERA-Interim reanalysis dataset (https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era-interim). Climate re-analysis is a process of assimilating observation data into a physics-based atmospheric model, a bit like Kalman filter (https://en.wikipedia.org/wiki/Kalman_filter) so you’d get more consistent state estimation of atmosphere across many different parameters, even ones that are not measured - here’s a good 3-min video on the topic.
Also, if you’re using the TMY file from Nice, note that this particular TMY is pieced together from 1982 - 1997 records by month by ASHRAE. I can’t say why the wind direction might be different but the methodology of piecing together month by month is obviously not going to be perfect and climate change is already affecting many places so do take any TMY data with a grain of salt.
For instance, you can see below how the highest temperature in Paris has gone up by 10 degC over the last several decades. In the TMY file for Paris, the maximum temperature is still only 30 degC.
