Since the hammer impact experiment is highly instationary you have to model this problem in the time domain (*ModalDynamic or *Dynamic in Standard or Explicit). Results should be passed through an FFT analyzer afterwards.
Random reponse is based on an ergodic, stationary excitation which...
C3D4 are very poor elements in strength analyses, I use them only when modelling masses. You need a very dense mesh to get accurate stress results. C3D4-Stiffness is also too high.
In terms of contact both element types work well.