The rating methodology on this website is exclusively test score based[1]. When I compare nearby schools, the school rating is almost perfectly inversely correlated with the % of Free/Discounted Lunch Recipients. The reduced lunch % is also pretty correlated with $/pupil, making the $/pupil inversely correlated with the lowest test scores, as you said.

But, that's like saying hospitals don't work because the highest expenditures are on the sickest people. This isn't measuring whether more money helps schools perform better, it's showing that we spend more money on schools with lower test scores.

[1] https://www.schooldigger.com/aboutrankingmethodology.aspx