I don't know whether you're doing anything wrong from a code point of view, but you're certainly not giving us enough information in your email. We don't know how you've got your results or how you've counted genes <0. Basically you need to show us the code you've used.
What genes are you refering to when you say they are "not identified"? Why should they have been?
Reading the help page for any edgeR function that performs statistical testing, such as decideTestsDGE, will tell you what cutoff values have been used.