A schematic view of a cross section is:
d(sigma)/d(p_T)d(eta) = ( N_{cand gamma} - (1-rej_pi) * eff_{cand. pi}* N_pi ) / ( eff_gamma* eff_{cand. gamma} )

This is missing things like sampled luminosity, which is needed to move from the pion cross section to N_pi.

Step 1:  _ Get N_pi.  These come from Werner Vogelsang's calculations
Step 2:  _ Get eff_{cand. gamma} and eff_{cand. pi}.  These come from running the gamma tree maker and comparing the candidate list to the list of all photons and pions in the MuDst.
Step 3:  _ Get eff_gamma and rej_pi.  These come from looking at the distribution of the discriminant for photons and pions above and below our cut.
Step 4:  _ Get N_{cand gamma}. Find and Run on the data to see how many gammas we see in the trees.

More on Step 1:
Jason Webb provided a text file with the cross section information in it, though there's some confusion abotu the actual units.  They are picobarns times some energy factor, depending on whether the variable is d(sigma)/dp_T or E³*d(sigma)/dp³.  If I can get that sorted out, I can use the sampled luminosity to compute the expected number of pions in each p_T bin, which is essentially N_pi.

Brief Update:  These numbers are in E*d³sigma/dp³ [pb/GeV²].