Regress out % mito RNA in snRNAseq OR remove mito genes from VariableFeatures

[tdaf]

Hello,

I am working with brain single-nuclei RNAseq dataset from a model organism, and trying to make decisions on the correct pre-processing/QC steps. I made some first crude filtering, keeping only nuclei with less than 5% mito RNA. Initially I kept regressing out "percent.mt" feature (module including all mito genes) during either in ScaleData or in SCTransform for both my whole dataset clustering, and specific class (GABAergic, glutamatergic) subset. It made the most sense since in a single NUCLEI sequencing dataset I would assume any mito RNA reads would be guaranteed artifact.
However, and please correct me if I am wrong, I understand that regressing out percent.mt might affect also any other genes that have correlated expression to that of mito genes? That becomes significant if I have cell types/states that I am trying to detect that might be inherently more metabolically active (which I suspect I do) and so express non-mitochondrial activity genes in a similar pattern to mito genes.
Should I keep regressing out percent.mt? Or can I remove all mito genes from VariableFeatures instead so I dont lose biologically significant signal from other metabolism-related genes?

Thank you for reading, and would appreciate any insight on this!