I believe there is no consensus on which is the best software. None of them are perfect, and as you can imagine is a vicious circle trying to determine which genes are stable across all your samples when there is so much variability introduced from RNA extraction to qPCR.
geNorm is the most widely used, but it does not mean it is the best. Then there is BestKeeper and Normfinder. Personally, I recommend you try all of them and see which genes are consistently the most stable. I also recommend you look at the raw Cq values. For me geNorm worked just fine.
Do you still have a pdf of the paper that Carina Kozmus attached. I'm am working on defining reference genes for an experiment and would be very interested in reading this paper. Which method did you end up using? Based on your question it sounds like your samples contained host RNA as well as bacterial RNA. If this is true, could you use NormFinder for your application?
I was working with RNA from hemocytes, without infection. I decided to follow Carina's advice and tried the three methods: NormFinder, BestKeeper and geNorm. Then, I selected those primers more consistent. Tipically, the three or at least, two of the software, agreed on the pair of primers more stable. Then, I selected that ones.