The second chart was likely thrown together in MS Excel, as evidenced by the format of the "12-Apr" release date values compared to the straight-text "Q2/Q3 2012" values. I suspect some NV fanboi was having fun in Excel, and made their personal wishlist of what they hoped Kepler would bring. Think about it this way: Who would make such a chart, and why? Who would want to compile this information? Not an NV employee. Someone in marketing might have access to such data, but would not format it this poorly and would not waste time on an internal-only document. Someone in tech might make up the price, release date and performance values, but would have more accurate technical details than are shown.
Perhaps a manuf. partner, but not likely given the adverse consequences from NV given a leak. And again, the mix of data in the table is suspect.
There are also some oddities in the content. Why copy the "PCIe3.0x16" all the way down if it's constant? Why does the 650 have 256 SPs instead of 384 (check ratios to bus, ROPs, etc.). Why put 2 GB on the GTX640? Why does bus width vary with SPs? It's not like each SP has its own dedicated memory, and the resulting increase in design costs for both the chips themselves and the board manufs isn't a sensible idea. I agree with most of you: it's bogus.