Should we filter RNAseq data based on the CV with risk of removing biological variation?

Question

I am working on RNAseq dataset. Some from mouse with very low biological variability between replicates which makes easy the conclusion that I should filter genes with high CV.<