Different methods of data normalisation produce different cluster markers

Question

[ria.dashyan]

I tried comparing different data normalisation methods in Seurat per my PI's request and ran into a problem. The SCTransform and the regulat NormalizeData workflow produce completely different DGE results. Which is more accurate and which one should i be using? I am analysing a dataset of scRNA-seq results with manually defined cell clusters based on the cell type because the pca and the FindClusters function can't accurately determine which cell type is which, could it be that the data given to me is just of such poor quality that nothing would work on it?