Difference in S-value labeling error?

[abio90]

I just wanted to report a potential mistake in the PrimateAI-3D webserver concerning the difference in S-values between primates and human. As I have understood, an S-value closer to zero means that the gene is under lower constraint (more likely to carry benign variants). However, when a gene shows an S(primate) < S(human) with a significant p-value, the website displays "Human less constrained in population model", when it should state the opposite sentence (and vice versa).

For example, one of the genes defined as more constrained in human in the the paper is TERT, however the webserver states "Human less constrained in population model". Isn't it wrong?
https://primateai3d.basespace.illumina.com/gene/TERT?dataset=gnomad_r3

Can you please confirm that this is an error? Or have I misinterpreted the S-value meaning?