Small question convergence criteria

Dear All
 
I'm modelling a beam strengthened with FRP NSM (near surface mounted) bar.
I did two simulation.
In a first one I use the following convergence method:
BEGIN ITERAT
   METHOD NEWTON REGULA
   MAXITE=250
   LINESE
   BEGIN CONVER
     ENERGY CONTIN TOLCON 1E-3
     FORCE  CONTIN TOLCON 1E-2
     DISPLA CONTIN TOLCON 1E-2
  END CONVER
(convergence governed by Displa) The last one is included in attachment
iN THIS CASE i HAVE A QUITE GOOD AGREEMENT BETWEEN THE MODEL AND EXPERIMENTAL RESULTS. (THE MODEL GIVE A lOAD-DISPLACEMENT CURVE A BIT MORE STIFF WITH RESPECT THAT OF EXPERIMENTAL).
 
iN THE SECOND ONE i USE THE FOLLOWING CONVERGENCE METHOD:
BEGIN ITERAT
   METHOD NEWTON REGULA
   MAXITE=250
   LINESE
   BEGIN CONVER
     ENERGY CONTIN TOLCON 1E-3
     FORCE  CONTIN TOLCON 1E-2
     DISPLA  CONTIN TOLCON 1E-3
  END CONVER
(convergence governed by energy)
In this case I receive an early debonding of the NSM FRP bar (the curve is perfect till the 80% of the failure load and 50% of displacement, after the load drop down)
Is normal a so big difference by changing the Convergence tolerance?
Can you let me know, please, which tolcon should be used for a good simulation?
Thanks for your help
Best Regards
Anipalm