You mean from 1976(/1996) to present? Yeah, I got that. My original comment was about why that's a pointlessly superficial way of looking at the data.
"Defense R&D includes a lot of basic science....for example, this paper of mine..."
...was mostly funded by the NSF, if the footnote is to be believed. That's not defense money.
In any case, I never said that defense research doesn't involve any science, I just said that most of it goes to defense contractors. I hate to break it to you, but the government didn't double defense R&D after 9/11 in order to increase the number of papers being written on antenna theory. Most of that money went toward things like airport bomb-sniffers and futuristic weapons systems. I'm sure you can find some pure-science papers that were written with that money, but that wasn't the point of the funding.
From about 2000 to 2010 it went from $100B to $150B and dropped to $140B after the "bloodletting".
Your "more nuanced interpretation" that "training of new scientists starting (~1996), and when they needed funding to really get their labs started (~2000), we stranded them" is simply wrong.
Read the graphs. We increased funding 50% in real (i.e. "constant FY 2012") dollars, in nominal dollars it increased marginally more (inflation wasn't that important). The bulk of the increase was in the NIH - look at the second graph which breaks it out by agency.
Defense R&D includes a lot of basic science, I have no idea where you get the idea it doesn't. For example, this paper of mine (http://arxiv.org/pdf/0708.0635.pdf) has support from the Air Force. A lot of work in computational electromagnetism and operations research (to name two fields I know moderately well) is funded by the military.