Top has proposed the following trigger for the 9.0 version of the trigger list (see their excellent documentation page).

Level 1	CEM(1,10)CJT(2,5)
Level 2	EM(0.85,10)2JET(10.)
Level 3	ELE(ELE_LOOSE_SH_T, 1, 15) Jet(SCJET_9, 2, 15)

A quick look at the efficiency for this trigger shows the following:

The reco cut is a reconstructed electron (EMPART_S) with ET>15, EMF>0.9, and Isolation < 0.2. If the reco pre-selection cut is not applied, then one observes the following (cuts are the same as above, just remove the "RECOELE" cut from the list).:

The substantial drop in the L3 electron requirement has to do with missing track, etc., information. Though this has not been checked, it is likely this is due to fiducual cuts.

Finally, if one breaks down the reco cut itself, one can see that much of the inefficiency comes from the 15 GeV and isolation requirement. The EMF requirement does little once the first two are applied.

Here is the reco pT of an electron as a function of the applied trigger cut. There doesn't seem to be a bias here, which suggest geometry. Looking at the Isolation and EMFrac one sees similar behavor.