This is like those polls that showed Obama tied in Tennessee with low 40s numbers, they just don't equal competitiveness in November. Let's look at some 2008 polls for Michigan.
Public Policy Polling September 6–7 Barack Obama 47% John McCain 46%
American Research Group Sept 16-19 Barack Obama 48% John McCain 46%
Marketing Resource Group
of Lansing Sept 15-19 Barack Obama 43% John McCain 46%
NBC/Mason-Dixon Sept 18-23 Barack Obama 46% John McCain 46%
Strategic Vision September 22–24 Barack Obama 48% John McCain 45%
I get your point about outlier polls like the recent Tennessee numbers, but don't think Michigan compares. The one thing that most of the numbers you posted have in common is that they all came in early or mid September 2008, the one point of that campaign where McCain was actually leading or tied with Obama. It was after the convention and Palin pick, but before the financial melt down. So McCain likely was close to Obama in Michigan during this period. Every poll out this month has shown Michigan a close race.
Even among that timeframe, he dishonestly hardcore cherry picked a handful of polls and ignored others. The RCP average from early or mid September 2008 in the daterange provided would have shown an Obama lead of +4.
Of course, now every poll is showing a tied race. No cherry picking necessary.
Because ignoring Rasmussen and PPP is in no way cherry-picking.
I'm not. The RCP average only uses the most recent polls.
As an aside, I've never known exactly how RCP decides what counts as "recent." It doesn't seem to be simply a certain number of polls, nor the polls conducted within a certain time frame.
In other words, letting RCP do your cherry-picking for you doesn't really change things.