The following are a collection of distributions at each level of the CMS muon trigger. Also included in these figures are pt resolutions and scatter plots of reconstructed quantities at each level plotted against the simulated quantites.

"Sanity Check" plots for Z->mu mu events: sample generated in CMSSW_2_1_8, analyzed using CMSSW_2_1_12 with IDEAL_V9 tag; baseline algorithm used for L3 reconstruction.
Note: the L3 phi distribution reflects a change in pixel geometry for -z from 2_1_8 to 2_1_12.
L3 phi distribution for muons with eta > 0.
Distributions for L2 muons that fail to seed an L3 (baseline L3 algorithm)
Distributions for L2 muons that seed an L3 (baseline L3 algorithm)
"Sanity Check" plots for Z->mu mu events: sample generated in CMSSW_2_1_8, analyzed using CMSSW_2_1_12 with IDEAL_V9 tag; outside-in state-based algorithm used for L3 reconstruction.
Distributions for L2 muons that fail to seed an L3 (outside-in state-based L3 algorithm)
Distributions for L2 muons that seed an L3 (outside-in state-based algorithm)

The following are from the Fall08 ttjets-MadGraph sample, generated in CMSSW_2_1_8, and analyzed in CMSSW_2_1_12 using the STARTUP_V5 tag:
Distributions for L2 muons that fail to seed an L3 (baseline L3 algorithm)
Distributions for L2 muons that seed an L3 (baseline L3 algorithm)
Distributions for L2 muons that fail to seed an L3 (outside-in state-based L3 algorithm)
Distributions for L2 muons that seed an L3 (outside-in state-based algorithm)

The following show the diagonal elements of the L2 muon error matrix used for determining the region used by the L3 reconstruction to find the matching tracker track. To improve the L3 reconstruction efficiency, different rescaling factors may be used. The following figures show the diagonal elements for Z->mu mu events
Error matrix elements before rescaling
Error matrix elements after rescaling

The following is a comparison of the CPU time for the module hltL3Muons for each of the seeding schemes for ttbar events and startup conditions (baseline is in black, outside-in hit-based is in red, outside-in state-based is in blue). This figure was produced using the tree utility; and the implementation of the module timing is still under development. The results should be considered preliminary.
Execution time for hltL3Muons in ttbar events

The following plots show the L3 reconstruction efficiency for the three different seeding schemes: inside-out hit-based, outside-in hit-based, and outside-in state-based. Algorithm efficiencies for Z->mu mu events, IDEAL conditions
Algorithm efficiencies for Z->mu mu events, STARTUP conditions
Algorithm efficiencies for ttbar events, IDEAL conditions
Algorithm efficiencies for ttbar events, STARTUP conditions

The following are trigger efficiencies for the HLT_Mu15 trigger path:
MLT_Mu15 trigger efficiencies for Z->mu mu events, IDEAL conditions
MLT_Mu15 trigger efficiencies for Z->mu mu events, STARTUP conditions
MLT_Mu15 trigger efficiencies for ttbar events, IDEAL conditions
MLT_Mu15 trigger efficiencies for ttbar events, STARTUP conditions
The following are trigger efficiencies for the HLT_Mu9 trigger path:
MLT_Mu9 trigger efficiencies for Z->mu mu events, IDEAL conditions
MLT_Mu9 trigger efficiencies for Z->mu mu events, STARTUP conditions
MLT_Mu9 trigger efficiencies for ttbar events, IDEAL conditions
MLT_Mu9 trigger efficiencies for ttbar events, STARTUP conditions