Those are great links. Thanks for posting them. But they appear to show the models almost exactly as bad as the the grandparent: both indicate reality is at the very bottom of the model prediction distribution. It's unfortunate that the grandparent is from such a sketchy source, as it demonstrates greater respect for the principles of visual display of information. It shows one thing, it shows it well, and axis that people care about (the vertical) is given reasonable scaling instead of being compressed away by cramming in multiple additional graphs.
Furthermore, consider the lameness of the first claim in the AR5 chapter you like: "Predictions for averages of temperature, over large regions of the planet and for the global mean, exhibit positive skill when verified against observations for forecast periods up to ten years"
This sounds good, until you realize that the best thing that can be said of the models' predictive capacity is that it is better than chance. That is what "positive skill means", and that is all it means.
As someone who has worked in predictive modelling, this is something that people only say when their model has no practical predictive utility. It is easy to get models that show results that are by any measure many standard deviations away from chance, but that are still completely useless for the kind of predictions required by policy makers. To take a trivial concrete example: a model that tells you to "drive east" when your destination is in fact in the eastern half-plane will give results that are far better than chance (which would be driving in any random direction) but it will only rarely get you anywhere close to where you want to go.
The report goes on to list a variety of positive results with varying confidence, but none of them add up to "predictively useful for policy makers" and that's for global and large-scale regional climate. Local climates--which are what we really care about--are far harder to predict.
This is not to say that models are bad science or "global warming is a hoax" or any such nonsense. There is fairly strong evidence that anthropogenic greenhouse gases are a significant contributor to climate change by adding 0.3% to Earth's heat budget at the surface, and that anthropogenic aerosols are likely removing about 0.1% of the effective insolation at the surface, for a net 0.2% gain. These conclusions come from observations on the ocean heat budget, the temporal distribution of warming (which is greater at night than in the day, for example, ruling out solar variation) and the geographic distribution of the warming. It's possible to say all of this--and so have fairly high confidence that humans are having a significant impact on Earth's climate--but still not have much of a clue how the highly non-linear climate system is going to respond in the near or long term.
In some ways, because our economic systems are relatively fine-tuned to the historical climate, which we can predict will undergo fairly significant variation even if we don't know precisely where or what, the details of the future climate matter less than the high confidence it is going to change. We should be investing in robust systems, or we will be facing a significant, ongoing global recession as climate conditions trash economic assumptions.
But claims like the one in TFA are necessarily strongly dependent on model details, and while it's an interesting study, it was done by climatologists, not computational physicists, and that shows in the excessive confidence they place in the detailed model results.