Cams irradiation time series single point product vs. gridded big difference

I have a problem with the gridded cams irradiation dataset. If I compare the values e.g. for lat=48.25, lon=12.25 (which should be exactly in the gridded dataset) with the single point time series product , I get for the yearly sum of GHI values in 2019:

gridded GHI data for point 48.25/12.25: 1.064 kWh/m2 (full year 2019)
direct single point time series via cams-API 48.25/12.15: 1267 kWh/m2 (full year 2019)

How can this difference explained? The values of the gridded timeseries appear to be generally (I compared different points in Europe) ~10-15% too low.

I would appreciate any help or support (Maybe I did something wrong?) - thank you.

K. Schulze