I guess we have to wait and see under real working conditions how unmodified LNBs would work.
I confess that my LNB is already modified (will edit this in my previous post) It has the original xtal inside but the plastic cover is broken in many places and the metal part exposed to the external world. This might a good explanation why I have so strong fluctuation up and down (+it's raining here)... Thanks for checking that!
I have a feel that might extra thermal insulation layer(s) of the LNB could drastically improve this. I learnt that from G0UPL Hans, when he shared his OCXO development experiences in case of WSPR mode: http://www.hanssummers.com/ocxosynth