Bizarrely, that link doesn't seem to give you any means of finding out what the methodology used to construct that ranking is - which makes me very suspicious of it. From this quote:
As expected there is a positive correlation between delivery of A*-B A levels in all subjects and delivery of high grades in STEM A levels, but the percentage of A level entries in STEM varies greatly per school and therefore so does overall STEM ranking.
I suppose it's combining how popular STEM subjects are at the school with how well students who take them do - which is OK if that's what floats your boat...
Eton's A level results can be inspected here. For example, in summer 2016, it seems 46 boys took A2 Chemistry, of whom 22 got A-star, and 43 got A-star to B inclusive. That seems, at any rate, good enough to be confident one's budding chemist will have the opportunity to do well if he puts in the work.
Maybe someone who has the statistics of one or more other schools on the list at their fingertips might like to reverse engineer what the algorithm must have been? (I might, but not tonight, and not honestly sure I care enough to do it at all!)