I have a bit of a problem with these stats myself being used as raw data like you are doing. Steam splits it at roughly 50/50 dual core, quad core, I can't honestly believe the majority of those buying games actively are not yet at 4 cores, given even at 4 cores I have problems running some games these days. I can however believe people adding their second work computer/laptop make up a portion of that statistic, those with inactive machines/accounts also, and finally those who bought one game once and never bought another.
I think it'd be helpful to cross reference the gpu stats and see how many of that 40% dual core stat are going to be gaming rigs, but I see no easy way to do it via the steam stats without manually adding up the upper parts of the GPU table (using those GPU's which can still run modern games).
Thats correct - publishers/developers etc would probably look more at what is currently being sold vs setups that are declining (see Wiz's comments about a month ago about how they are not going to test on single core PCs notwithstanding problems may appear). There is, for example, a known issue that laptops are a significant part of the 'gaming' market, so the (not terribly developer friendly) intel chips form the bottom end of what games should perform on.
Another issue with the steam stats is that the vast majority of accounts have only free to play games on them (assuming trading/spam accounts tend not to trip the survey) - most of which are probably not terribly intensive so can easily run on old hardware.