Going through all I read on this forum I decided to check everything related to my project. I stared by weighting my tool which is 1.3 kg and I calculated the center of mass from SolidWorks. On the other hand I read the robot external forces at the TCP and get the inaccuracies on each axis and apply the offset to the threshold I'm using for my measurements. It seems to be working good when I'm detection shear force at 5N, and no damage was caused to my sheet metal part.
What I want to do next is : if even using the offset of the inaccuracy and I still have a non valid force I'll switch to use Seulki's method that uses jointTorqueCondition.
Thanks everyone