Actually, this appears to be the result o f 90 runs of one model, which is not at all the same as "95% of models", it is more accurately "95% of 90 runs of a single model". This is typical of labeling done by people with an agenda.
The models used in climate research are first-principles physics models. After the parameters and relationships in a model are set, that model is then run repeatedly to get an average for that model (Monte Carlo simulations). A more accurate chart would show JUST the black average line from observations (which are statistical estimates), along with averages from competing physics-based models. Such a chart is available.
By the way, any statistically based time series models are pretty much meaningless except as ways to summarize observations of data, as they ARE able to be tuned (using linear regression techniques, leading to an error called over-fitting). I mention this as an aside, since Spencer is apparently correctly using the physics-based CMIP5 model.
I know that many will probably dismiss the IPCC summary figure below, but it is what a chart of multiple models would look like, especially note the legend listing all the models used. I don't like IPCC for their politics, but their science is more trustworthy.
Of course, since I am a life-long analyst and only a recent politician, I would point out in Figure (a) that the models appear to be drifting high over the last decade (2000-2010). I would have asked my graduate school math students to explain this drift, or go back and figure out what they were not modeling.
But, as I say over and over, I am fighting the "anti-regulatory war", not the "global warming war". And it is that fight for freedom that drives me.