Under advice from someone here I went to the ANL site and then the report at
http://transportation.anl.gov/pdfs/TA/559.pdf
This is a recent and thorough study that works out the CO2 and petroleum saved with conclusions shown in Fig. 24. This is no easy read, but it clearly states that the "Regular HEV" causes GHG Emissions that are about the same as the PHEV40 charging from the Illinois mix and and slightly higher if that PHEV40 charges from the US Ave. mix. The difference is greater if they assume the California mix. I disagree with thier idea of marginal response. EPRI also disagrees. I disagree with EPRI as well. But the differences are not terribly important. Ignoring my opinion, the gain from making a hybrid into a plug-in are still not all that much.
ANL persists in reverting to the wrong question in their conclusion, by comparing the plug-in hybrid results with the conventional car results. Thus they can make the plug-in seem like a big improvement. They never address the relationship with the hybrid that nobody fooled around with.