When LST is used for seeding, there is an efficiency drop in the OT-IT transition region (2.5 < |η| < 3.0). This drop is not present when seeds not duplicate cleaned by LST are used for the seeding. This effect is shown in the plot below:
This seems to imply that the pLS-(any other object TC) duplicate cleaning kills pLSs that could be extended to a true track, while the remaining TC is not a good reconstruction of the true track and gets dropped downstream. One way to test this hypothesis would be to disable the pLS duplicate cleaning for |η| > 2.5 and see how the performance changes.
When LST is used for seeding, there is an efficiency drop in the OT-IT transition region (2.5 < |η| < 3.0). This drop is not present when seeds not duplicate cleaned by LST are used for the seeding. This effect is shown in the plot below:
This seems to imply that the pLS-(any other object TC) duplicate cleaning kills pLSs that could be extended to a true track, while the remaining TC is not a good reconstruction of the true track and gets dropped downstream. One way to test this hypothesis would be to disable the pLS duplicate cleaning for |η| > 2.5 and see how the performance changes.