PPP is using a likely voter model that doesn't weight for party ID (a fact which they are very upfront about). Comparing the party turnout in this poll to past exit polls is comparing apples and oranges - a likely voter model in August ≠ final turnout in November.
An argument that this poll is bad news for Akin is legitimate (I think it is bad news because the likely voter model will probably change between now and November) but to accuse PPP of intentionally messing with the numbers shows a lack of understanding about what this poll is showing.
Is what it is "showing" just a bad methodology? And just why is this outfit with a somewhat Dem house effect, suddenly apparently crossing the River Styx? I have asked these questions. I am not yet satisfied with the answers so far.
It's only a bad methodology if what you're trying to see is what they "think" the results will be in November. PPP could weight these numbers to what they "think" the partisan makeup of the electorate will be in November. But that's not polling; that's electoral projection (also known as guessing) based on some data points.
If what you're trying to see is the current margin based upon voters who say today they're likely to vote in November - which is all this poll is - then it's exactly what PPP is offering.
As for why PPP's sample in particular got so much more Republican from their last poll of Missouri, this is the first time they've used a LV model in Missouri (their last poll was of RV's).