After previous difficulties I rebooted and started over. 

Changes from Will's code, only in StEEmcMixMaker.cxx

1)Comment out the same sector cut:

    /// same sector
//     if ( point1.sector() != point2.sector() )
//       {
//         numberofFailpoint.push_back(1);
//         continue;
//       }

2)Add an energy-dependent opening angle cut.  I get the momentum vectors of the 2 clusters:

    TVector3 momentum1 = (point1.position() - vertex).Unit();
    momentum1*=point1.energy();

    TVector3 momentum2 = (point2.position() - vertex).Unit();
    momentum2*=point2.energy();

   and the angle in between them:

    double deltaPhi = momentum1.Angle(momentum2);

   then sum together their energies:

    double totalEnergy = point1.energy()+point2.energy();

    now define the angle bounds based on energy:  

    double minAngle = 999.;
    double maxAngle = -999.;
    //so i get the min by calculating the angle for zgg=0 and the max for zgg=7 and subtracting/adding 25% from the min and 50% to the max
    if(totalEnergy<=10){
      minAngle = 0.0825;//zgg=0 Phi=0.11
      maxAngle = 0.225;//zgg=0.7 Phi=0.15;
     
    }
    else if(totalEnergy<=15){
      minAngle = 0.055;//zgg=0 Phi=0.073
      maxAngle = 0.15;//zgg=0.7 Phi=0.1;
     
    }
    else{
      minAngle = 0.033;//0.044
      maxAngle = 0.1155;//0.077
    }
    std::cout<<"total Energy = "<<totalEnergy<<" phi bounds "<<minAngle<<" to "<<maxAngle<<" with this phi "<<deltaPhi<<std::endl;
    if(deltaPhi< minAngle || deltaPhi > maxAngle)continue;

In summary, I don't look at which sector they are in at all.  So I presumably have a mix of same-sector and neighboring-sectors.

Some plots are below.  Note that everything that says "pi0" should really be "diphoton" and that "versus mass" is an artifact of the old 2-d histograms from which I projected these.  All plots show the entire invariant mass range in the figure on the upper left.  I have only run over a small fraction of the data to get these and I've got good stats already.

What's next?

Once everyone agrees that my events look good, then I'll process the rest of the data.  I need to hear from Will & Ilya about a techinal question for combining the shower shape trees.

Update Tuesday June 16 - Merging tree difficulty seems to be resolved.  Now just working on including the useful information from the pion trees with the shower shape trees to make QA easier/better.  I cleaned up a bunch of stuff in the pion code and changed to save all trees in one file.  Now just testing and debugging this.  I had some trouble when NFS went down and did not get as far as I had hoped I would.

In the meeting this morning we discussed some ideas about the phi distribution spikiness.  We all agree that some spikiness is a going to be seen from the sector boundaries, but perhaps right now it is still too big.  One idea is to merge points across boundaries in some radius (~2.5-5cm).  Or make a fiducial cut.  Maybe the 2nd photon in an eta near a boundary is getting split into 2.