If you want look up Elowitz M on pubmed, it was in a talk he gave. He was giving a talk on noise in transcription, which is obviously highest when dealing with low copy numbers where stochasticity leads to larger variance in levels. It was obviously an extreme case and I don't know where to track down a reference and finding that kind on info in a lit search is a pain. I know what a half-life is. When you give a steady state level, you need to know the decay rate to make a guess about the production rate. So to comment on how many were made an hour, I had to comment on the decay rate. My primary point is best made in the context of selection. It would require infinite work to get a completely error free process even at the phenotypic level (which would be more buffered). So phenotypic effects will happen at some level due to mutations in mRNA (maybe only partly). I already stated that it wasn't an important point anyways so I don't know why you're being so antagonistic.