Fix issue with JMS calibration
This MR fixes a bug that was introduced in !42190 (merged) when migrating the large-R JMS in situ calibration to master. The issue was reported in the AMG meeting. I simply declared the same TLorentzVector twice, once within an if statement such that the TLV wasn't filled properly outside that criteria anymore. Code has been tested locally and runs fine now and the output looks as expected.