As much as I like the graphs, I think there is one important thing missing: the times add up if you employ more than one of these methods, which is often the case and often required for different stakeholders(*).
Sure, I sketch a lot, but I also build interactive prototypes for the interesting bits, and I have someone spend time on making it look good (i.e., adding unnecessary fidelity) too for a presentation-and-feedback session with executives.
So I am doing sketches + mockups + interactive prototype.
As you indicate in the fourth graph, you can save time by not doing real code, until you have to do real code anyway. And this is true for all of the fidelity levels.
Don’t you run into situations where you have to do more than one level of fidelity to reach the right confidence level?
(*) And yes, I realize that at Basecamp you are all developers, users and decision makers of the software, so the types of stakeholders may be limited compared to agency or other in-house contexts.