Some questions about propertion of co expression network

[HelloWorldLTY]

Hi, I compare the output of sctransform and sclink for gene expression network inference, and I found that the number of such link (edge) is very different. Consider I have 1000 genes which are used to find the correlation matrix.

For sclink, I nomrally have 1554/1000^2 = 0.001554, but if I use sctransform and p value as 0.005, then the propertion is 0.12443. It seems that sclink generally captures less co expression relation. Is it normal? Thanks a lot.