Honestly the charts in the paper is not very convincing. The outliers are all over the place and there is hardly an association judging from naked eye.

(e.g: imagine seeing figure 4 without the model prediction in the background)