Note that it's not a 35% improvement/decrease, it's a 65% decrease, or put another way, 1/3 the emissions. It's not an order of magnitude, but a factor of 3 ain't bad, and CO2 from battery pack production will get better as the energy sources used get cleaner. Also, it looks like they're assuming a car lifetime of 250k km, which I think is pretty conservative for these. They also assumed a linear decrease in CO2 in power generation, which they say is a conservative assumption, and I agree. A number of their other assumptions lean on the conservative side, I assume for being more defensible, so I'd be somewhat surprised if the real difference isn't a good bit larger.