How can I eliminate low particles recording at histograms?

I made a simulation of a scintillation detectors. I got Edep from isotropic source of 662 keV, but I have a peak near to zero. I do not why, but in real data it does not have sense. Does anyone know how to solve the problem?

maybe these are steps that are non-physical, e.g., crossing the boundary of your scintillator material?
In example B2, they filter for deposited energy > 0:

G4bool B2TrackerSD::ProcessHits(G4Step* aStep,
  // energy deposit
  G4double edep = aStep->GetTotalEnergyDeposit();

  if (edep==0.) return false;

@weller I tried to do that, but it gives me an error. I had to do that in Event Action class because I am saving edep through user action classes. Thank you very much for your help!

return false aborts execution of the current step (and G4bool is probably the wrong type also), so in this case maybe it helps to invert the condition?

if (edep > 0)
    <code to add edep to histogram>

what I made and really helped me was in Event Action class. I think if i made in Stepping, it would not be correct.
void myeventaction::EndOfEventAction (){

if (Edep==0.0*MeV) return
man->FillDNtuple (0,0,Edep)

Where Edep is an integration of every edep collected in each event.