I couldn't ever overlook the gross misunderstanding of statistics here. The "bell curve" is an asymptotic distribution law, and even assuming that the distribution of performance is indeed normal, it is outright idiotic to fit the bell curve with as low as 10 people. The chance that no one out of 10 random people will be really at the lowest 5% of performance, is as high as 60%. The HR even uses an Excel sheet to even draw a normal distribution...
I have seen many good people's careers stopped in my farm (and it is a big farm), because of this "bell curve".