Initial feedback via an email from my wife:
The document was published by an individual (Yoshiharu Nakano) who refs, upper left, an US Military (something like USN Air Information Bureau or Service) source which published revised data in '46. original data in '45. So, the ultimate source is some sort of US military eval that released this data some time near the end of the war in a publication: Japanese Aircraft Performance and Characteristics (revised 11/45 report is of data for Ha-chi-tatsu (? not sure) from US Test data - apparently they had one and tested it some time prior to publication). Speed was found to be faster by the US testers than in the data from the Japanese publication - @5460m 568kph (20-30 kph faster than Japanese data, but I don't have source on this comparative set). There's also some comment that the data is unusual in that it is for "combat" (what she calls "fighting"), which I suspect is like WEP or something like full mil.
So, Krusty's guess appears to be a good one - it looks like original source here is US eval and not Japanese.
My wife did that hurriedly and away from her tube. I told her to find me an attribution. She says that appears upper left - you can see the Romajji dates there.
As for the box lower left, it is a simple legend. She came back with: "The upper left graphs show all outputs. in the upper right and lower left, - and -.-. indicate military outputs. .... and -..-.. are fighting outputs."
I'll milk her on this some more, if you llike but I think that source info is what you were after. As for the speed discrepancy, perhaps the US evaluators were using better US Tar..? or maybe just a high-quality sample, fit and finish-wise..?