In the past Monte-Carlo simulations have been used to study the influence of morphological properties of fractal surfaces on the selectivity between competing three-step catalytic reactions. These simulations were conducted using abstract two-dimensional diffusion-limited aggregate models. In an attempt to identify the appropriate properties of a real material that should be represented in a model this previous work has been extended to consider three-dimensional models (cluster-cluster aggregates and the Menger sponge), in order to correlate selectivity with model characteristics which may be related to input from experimental characterisation data, such as that from magnetic resonance imaging. The simulations on three-dimensional models have shown that both the porosity and the mass fractal dimension of the models are important to accurately represent a real porous material; a finding not explicit in previous work in two dimensions.