WMAP have a bunch of per pixel model fit files available at Lambda. Sergi (and others?) have looked at these before. In the first PRL paper we quote a mean sync spectral index from "model f" ('fs' in filename) within our field as -3.3. I heard from Lyman Page and Raphael Flauger at the meeting in Minneapolis a few weeks ago that "we get -3.1" so I thought I'd take a look myself. The plots below show the results. The filename is given at the top of each plot. The file has three columns as labeled. At right we zoom in on the BK field and apply a binary mask to the non-zero coverage region. There are three histograms in each case: full sky, binary mask applied, and pol apodization mask weighting applied -the title of the histogram gives the weighted mean value for the last case.
There are two other model where the sync spectral index is a free parameter: "model c" ('base' in filename') and "model g" ('fss' in filename). We see below that "model c" gives very similar results. "model g" gives a slightly more negative value although I am not sure what this means as it is apparently not a simple power law index.
