G4GenericBiasingPhysics unexpectedly increases hits in sensitive detectors

Hi all,
I have implemented biasing in my project to split particles entering a specific volume (100% non physical bias). After some testing, I found out that even if I wasn’t doing anything in my biasing classes the number of hits in the sensitive detectors were higher and the energy deposited by each hit on average lower (total energy deposited stayed approximately constant). After a bit more of testing, I discovered that this behaviour is solely due to the following command:
physicsList->RegisterPhysics(new G4GenericBiasingPhysics()). No need to specify which particles to bias, nor to attach a biasing operator to any volume. I was not expecting this to happen and I would like my code to still produce the same physics in all the volumes unless I specify which volume and which particle to bias. How can I obtain this? I have reproduced this behaviour in the example B5 by simply editing exampleB5.cc and running more events. The edited files are attached. I am running Geant4 version 11.0.2.
Thanks for your help
exampleB5.cc (4.1 KB)
run3.mac.txt (557 Bytes)

Hi all,

here is an update. It is not actually trivial to observe the effect of the biasing class when using the default physics list of example B5 FTFP_BERT. I have compared the number of hits in the calorimeter, the single hit energy, and the total energy in the calorimeter (See pictures FTFP_EC_* where black is bias on and red is bias off). I am putting a link because I am allowed to only upload one picture:
https://drive.google.com/drive/folders/12vKra4yNkJkiLa83hwFDDFq4QatXmv2P?usp=sharing

The effect is exacerbated when using a more detailed EM physics list (G4EmStandardPhysics_option4). In this case I observe 10 times more hits when the GenericBiasing class is registered but doing nothing (See pictures Opt4_EC_* where black is bias on and red is bias off). Each hit has on average less energy so that the total is more or less similar.

This difference is in my opinion very baffling and it has an enough important effect that I cannot use biasing in my project. You can reproduce my results copying the edited B5 example from the link and running on the macro run5.mac. To run with activated bias just remove comment marks from lines 81 and 83 in exampleB5.cc.
run5.mac.txt (553 Bytes)