Because that image, as it is, is very very frightening. If it holds up to more investigation, there's something terribly wrong -- there should *not* be any correlation between exit polls being wildly off and what kind of machinery is used. On the other hand, I'm willing to also accept that someone is just posting the data they want us to see, leaving out the wildly off for paper ballots and the closely aligned for electronic.
Your duty, should you choose to accept it, is to give me more info.... :)
I found more information in this Slate article, thanks to inmostlight. The article itself unfortunately doesn't give its own graphs of exit polls vs actual voting per state, it just gives a list of states it says weren't so off. But I did compare this map to the map I posted earlier -- the new map authoritatively gives which states (and if you go further, which counties) use which sort of voting machines. And, with the exception of Florida and New Hampshire, all of the states which are listed as "Electronic Voting" are in fact mixed voting machines. In the case of Florida, it is all electronic voting... but in the case of New Hampshire, it's all paper ballot.
So that solves that mystery.