Jáchym, the averaging method you mentioned (Tmax + Tmin / 2) is the one that NOAA uses, I believe, to calculate mean temperature. I'm not even completely convinced that's WU's method. I've occasionally noticed very small errors in their mean temp (like two or three tenths of a degree off) using the above method as reference. I agree, it would be better to take the raw values and average them ALL.
You won't hear any arguments from me against your other points. I will add that CWOP (NWS citizens network) does the same thing -- punishes you in data quality scores for what your neighbors are doing. I've gone to great lengths to site and maintain my equipment as best I can only to be told that I am off because my neighbors have old sensors, didn't bother to put in their elevation in the barometer setup or put their anemometer near ground level in a sheltered area. That stuff drives me nuts. At minimum, there should be a form to communicate your siting details to any provider. Best case scenario, there should be humans with basic training in siting that can review your site (using the form data and pictures) and make a judgement call as to how you get weighted in the quality algorithm. Of course, this assumes competent, objective people who can 1) train others and 2) receive such training. These days, you might as well ask a bear to play golf.