When a particle experiences a nuclear interaction while being channeled by a crystal collimator, a variable "t" is used to calculate the resulting change in angular coordinates. This variable was not initialized when the particle was absorbed (i.e. lost) in the interaction. While physically there is not any "angular kick" to speak of since the particle is lost, the missing initialization could lead to different deflection calculation depending on the building environment, with the crystal-related tests failing as a result. This pull requests fixes this issue by adding the missing variable initialization.