pearson residue normalization project to tSNE coordinates vs UMAP and the surface outline property?

[brainfo]

Hey, I really don't know if this could be called an issue.
Since in the paper justifying square root normalization for visualization and clustering? and the Pearson residue normalizes and selected hvgs, both of the results are projected onto tSNE space. Are you suggesting we use tSNE as a major distance projection method?

I tried with my data and it seems with the UMAP projection, my data distribution surface seems less differentiable than other normalization methods. And I check in Pearson residue in this paper Fig4d,e you also see more "linear" /less differentiable distribution on the tSNE embeddings (compared to fig4a and c).
Actually, I personally like the simplicity of linearity but I reckon I need some explanation with regard to how this is the case (e.g. to explain to my colleagues and such since they are using sctransform and get the projection as in fig4c) and the confirmation that tSNE embeddings outperform UMAP with pearson residues.

Best,