I would dearly love to know where those plots came from. I once found them on Wikipedia, attributed to me! Where did you find them this time?
Anyway, take a good look. The first thing is that the categories are mostly wrong. Most states had a mixture of technologies, and only Maine had a substantial proportion of handcounted paper ballot precincts. It did, however, have a very small exit poll discrepancy ("redshift").
Of the other two that are supposed to have "paper ballots", Illinois appears to have been mostly optically scanned ballots, which are indeed "paper" - but counted by machine. And contrary to the plot, it had a 4.8 point redshift. Wisconsin was mostly optical scans - and it had a 5.3 point redshift, again, contrary to the plot.
Of the remaining six, all listed as "Electronic voting": North Carolina certainly had a substantial number of DRE precincts in addition to optically scanned precincts. Its redshift was 8.8 points. But New Hampshire was mostly optical scans with a few HCPBs (i.e. similar to Wisconsin), and had a massive redshift of 13.6 points - but a clean recount. New Mexico was mostly DREs, and had a redshift about the size of Wisconsin's (5 points). Florida was half DRE and half optical scan, and had a redshift of 3.9 points (less than Wisconsin or Illinois). Ohio only had one county with electronic voting, as you know - the rest were mostly punchcards, with a few optical scans. It had a substantial redshift of 8.6 points. Pennsylvania had some DREs but more levers, and a more substantial redshift (11.5).
Most states are not shown of course. Of note are Texas, which had a substantial proportion of DREs but a net
blueshift of 4.1 points (larger than Florida's redshift) Colorado and Tennessee also had a substantial proportion of DRE precincts, but were blue shifted relative to the exit polls. New York, which had a huge redshift of 13.9 points, but was conducted on levers, with which New Yorkers seem extremely happy. Connecticut also had levers and a subtantial redshift (9.4 points). Delaware, however, had touch screens and a very large redshift (16 points).
In other words, those plots do not give the correct voting technologies; they do not give the correct exit poll discrepancies and they are not representative. There is, in fact, very little that can be concluded from the correct data. If all you had were Delaware and Maine, you might want to conclude that fraud was perpetrated on DREs. However, if you had New York, Connecticut and Pennsylvania, together with Texas, Tennessee and Colorado, you might want to trade in your levers for DREs. Unless you were a Republican.
I do wish those plots would disappear from the internet. They are simply wrong. Someone needs to put them out of their misery. I don't know where the exit poll discrepancies came from (I got my figures from the Best Geo estimates in the Edison-Mitofsky report - the estimates made on the basis of exit poll responses alone), nor do I know where the state machinery data came from, but it is wrong too. I actually got my data from the NEP dataset, but you could check it out here:
http://electionline.org/Default.aspx?tabid=1099although it's been updated.