That's about where I've stood for over 10 years. What I've refined since then is my appreciation of weather which controls climate. The typical Trenberth diagram showing annual net energy flows including back radiation from GHGs is basically useless since changes in water vapor on an hour by hour basis change the numbers. Not to mention seasons, diurnal, cyclical, etc. Anyone saying that backradiation will increase by X with "water vapor feedback" deserves a big "yeah right".
The most important thing to remember is that while the energy exchanged between earth and space (solar in, heat out) is in equilibrium over the long run (which means that increases in GHG cause warming), that energy flow equilibrium means jack for water vapor. Water vapor will always concentrate unevenly and cause more warming here and less there. Without knowing that distribution (i.e. the unevennes of WV), there is no way to estimate the sensitivity. It is completely incorrect to use the energy flow equilibrium to postulate constant RH.
I guess I should mention models: GIGO. They don't depict convection in enough detail to know what latent heat is transfered to the upper atmosphere and what the all-important distribution of water vapor is. Find me a model that does this: http://weather.unisys.com/surface/sfc_con_dewp.html accurately. Or even more importantly, this: http://www.ssec.wisc.edu/data/east/latest_eastwv.jpg
Strongly suggest that you read Roy W. Spencer’s new book “The Great Global Warming Blunder”. He makes all these points and more, and gives evidence that the AGW types have confused “cause” and “effect” on the feedbacks in their models. He specifically labels the “water vapor/cloud cover” cycle as being the main controlling negative feedback factor controlling global temperatures (the AGW types assume that said feedback factor is positive rather than negative).
There are models that handle water vapor with some accuracy. They’re called weather models. They require fast supercomputers, and have some value for 3-14 days.
Weather is inherently chaotic. Therefore climate is inherently chaotic. Changes in CO2 must be correctly modeled on a daily and yearly basis before the model can be valid.
The so-called Global Climate Models (GCM) oversimplify, and cannot have the sensitivity needed to predict climate. Albedo changes (caused by changes in cloud cover at different altitudes) are critical. Ocean current changes are critical. The GCM do not model any of this in enough detail to pick up chaotic changes and their relative probability. In simple terms, Garbage In, Garbage Out (GIGO).
A good model would predict El Nino/La Nina events and their duration. A good model would have the sunspot cycle and its relation to cosmic rays, which affects cloud formation, rainfall, and albedo. A good model would have random generation of volcanoes.
A good model would have decreased high-sulfur coal burning in Russia and Eastern Europe, starting in 1989, which led to less cloud cover in the Northern Hemisphere , which led to temperatures almost as high as 1934 in 1998. A good model would have the Urban Heat Island effect, and its effects on the thermometer record in urban and rural areas.
The current GCMs, as designed by the Climate “Science” community, have none of this.