Well first off I mostly ignore polls and blogs except to look at their sampling rates and crosstabs if time permits, it really didn't so very little reconciling happened. It might help if I explain how I do polling locally and if you want I can even email you the excel sheet I juggle numbers on the downticket with, but loosely:

Q1: "Would you describe... Dem, Ind, GOP"
Q#: "In the race for X between Republican John Jackson and Democrat Jack Johnson who..."

And that generates a grid, a 3x3 matrix in this case.

A B C
D E F
G H I

Where A for instance is self-identified DEms voting for the Dem candidate and E are self-ID Indies who are undecided and I self-ID GOP voting Republican. I'm not really interested here in the sum of responses, just the percentage for Self-ID and that candidate.

Now I say, "These represent the average voter in outlook, but not turnout figures", let us assume X+Y+Z=1 and X is the percent of the voters who Self-ID as Dem and Y as Indy and Z as Republican.

DEm John Jackson get: Dem = AX + DY + GZ
and Republican Jack Johnson gets: GOP = CX + FY + IZ
and Undeicdeds are Und = BX + EY + HZ

and incidentally those C and G values, Dems voting Republican and Republicans voting for a Dem, are decently high even on up ticket candidates but when I tell a far-downticket candidate "You need to consolidate your base" I just got done circling G and I and showing him or her some of their colleagues lower and higher scores respectively.

Now that's all pretty ironclad, as much as any polling can be, but then to get a real projection we have to pick values for X, Y, and Z and guess how UN is going to break. We use historical data for that in part but places change a lot in even 20 years and you'll only have 4 or 5 races to compare with. So it's a guessing game and I'll pick a best and worst case reasonable scenario... ironically my worst case scenario nailed every local race on the head within 2 points except a 3-way race. I remember telling one of them when asked how to counter that, "You can't, if they pull of that turnout we get crushed, end story, but they won't" of course in my own head 'they won't' actually meant 'no point worrying about it since we can't do shit if they do'. Well they did. So the polling data was right but the projection for turnout wasn't and polling that is all guesswork I don't care what anyone says, ten educated guesses, 1 will be right. I've never seen a likely voter modeling system that 'made sense' and I really do think everyone with reputations for getting it right just guessed mostly right twice in a row and did some spin and convincing to make them sound righter.

So not really blindsided, at least locally, but then Romney polled very well with Republicans locally and they voted that way too, it was the R-D-I make up on Nov 6th that blind-sided us and it's not like no one believed it possible, just thought it unlikely and knew there wasn't shit to do about it if it was.