You should refrain from speculating on highly specialized and technical topics because it is a non-scientific approach. And since you are not a researcher who actively creates new AESA radars, where did you get that 80% number from?
The word 'air cooled' is not very well defined in itself. For example, Leonardo calls its 'osprey' range of radars as air cooled, but they are used as surveillance radars. As far as I know, there is currently no concept of an 'air cooled' radar for fighter jets. I can think of only one explanation for the Chinese claims of air cooled radar for JF-17: a SIC substrate layer for efficient thermal dissipation.
Given the same total input power, i.e., the power that is used to generate radar waves + the power that is needed for cooling, and the same semi-conductor material for the TRMs, an air cooled AESA radar will perform better in terms of power conversion than a liquid cooled AESA radar IF boundary layer defects between the substrate and semi-conducting layer do not play a role. There is a fundamental underlying assumption here: the air cooled AESA uses materials having more efficient thermal conductivity, which is the case for SiC substrate. This superiority of air cooled AESA will be true up till the point where the substrate layer can no longer efficiently dissipate heat. If you go beyond this limit, the air cooled radar cannot keep up, and you enter the realm of liquid cooled AESA radars.
But in the previous paragraph, I used the word 'total input power'. The entire paragraph is written from the perspective of total input power. It is easy to overlook the assumption that both the air-cooled and liquid cooled version have exactly the same semi-conductor material in the TRMs. The total output power per TRM is determined by the bandgap of the semi-conductor, and so you won't see any difference in terms of output power. Which makes your 80% number completely ridiculous and shows you have no understanding of the topic on which you are opinionating.
A final subtlety is that in paragrpah 3 I am talking about going beyond the thermal capability of the substrate, and in paragraph 4 I am saying that the bandgap determines total output power. The TRMs are based on transistors working as power amplifiers and the manufacturing process limits the amount of power amplification you can get. So, if you keep improving your manufacturing, in theory you could go beyond the ability of the substrate. Another way is to increase the total input power, which again is limited by the manufacturing process. This is currently an area of active research and nothing can be said with surety, until the Chinese publish their research.