We consider the relation between the microscopic and effective descriptions of the unfolding experiment on a model polypeptide. We evaluate the probability distribution function of the work performed by Monte Carlo simulations and compare it with that obtained by evaluating the work distribution generating function on an effective Brownian motion model tailored to reproduce exactly the equilibrium properties. The agreement is satisfactory for fast protocols, but deteriorates for slower ones, hinting at the existence of processes on several time scales even in such a simple system.