I use RSEM to align and quantify RNA-seqs and then use edgeR to do the differential gene expression analysis with or without ERCC normalization.
The problem is that I almost always see pairs of "wings" at the bottom of the -log(P-value) ~ log(FC) plot. For instance,
At the bottom of the plot, you can see a set of genes on the left and another somewhat symmetric set of genes on the right are obviously separated from the other genes. I have seen such a pattern over and over again, with or without ERCC normalization before feeding the expression data to edgeR.