The two attached .pdf files show the results of a preliminary set of endcap cuts including the SMD.  In "Profile by Energy Bins" the energy bin is selected in the third cut, and the interesting cuts begin after cut 5.  The SMD cuts occur as 10,11,13,14,and 15 (12 is a sanity check making sure that the SMD centroid is reasonably close to the tower centroid, which it is.)

10 and 11 insure that the energy in the preshower layer is a certain fraction of the total tower energy.  This is a rough way of checking that the shower has developed to the extent that it should for a single photon.

13 checks the amount of energy to one side of the SMD peak (whichever side has more energy) as a fraction of the energy in the SMD peak itself.  IF the sum of these fractions in the two planes is too high, it is likely that there are two clusters, and the candidate is cut.  I may separate this out into a per-plane version.

14 and 15 check the amount of energy on the high side of the SMD peak in each plane, and cut if that value is too high.  This is similar to cut 13, but does not normalize by the energy under the peak.

In "Cuts and Profile", more detailed looks at the reasoning for these cuts can be seen.  Plots are shown divided into energy bins of 5GeV each to show how the various SMD variables change as a function of the tower energy.  A better clustering algorithm (I am currently using a homemade one) might extend the reach higher, but sensitivity to fluctuations and tails is likely to be a much mroe important factor.