On the Computation of Convolution in Actuarial Problems. Some Further Results