All the numbers I reported are from papers on pubmed and stuff, most of which are linked to from the wikipedia article. I spent several hours reading and re-reading papers; I didn't cite them because my normal blog audience doesn't care. But yeah, you can verify everything by starting out on http://en.wikipedia.org/wiki/Caffeine
Attention fluctuation -- look up biphasic sleep, or studies on the effectiveness of napping.
As for polyphasic sleep and napping, it is known that energy levels fluctuate cyclically throughout the day with two peaks, but I'm surprised that you plotted it with a sinusoidal curve that would suggest constant periods and surprisingly, constant amplitudes... of course, without axes the graphs don't say much of anything, but it's still a surprising presentation. You could justify it as being for a general audience, but it has significant risk of creating misconceptions, and weakens this article quite a bit.
By the way, one of the papers finds "Caffeine consumption for 4 weeks also significantly reduced hippocampal neurogenesis compared to controls." Which is not happy news for me.
It has day-by-day descriptions of symptoms etc. I was back to normal faster than expected, but I used certain strategies to help get there.
Enjoyable nonetheless. Is it possible to get some sources? In particular, 2-3 weeks for tolerance build up, and 5 days to "adenosine normality"? 5 days is a surprisingly small figure, and I would guess there is a rebound effect if it's just a short relapse. No expert here though, but coffee is my buddy so I want to care :)
Actually, would it also be possible to get a source regarding the attentional fluctuation throughout the day?