On the origin of the width of the ferromagnetic resonance curve