PoreXpert is an inverse modeller - it produces a stochastic range of void structures, all of which could replicate the percolation characteristic data input. Therefore, as explained previously, there are an infinite range of possible void structures that could generate the percolation characteristics of your sample. PoreXpert uses a series of approximations to produce its simulated void networks, so PoreXpert is not particularly good at producing absolute values of pore-fluid properties. Its strength lies in looking at a series of samples - for example a series of sintered filters manufactured by sintering particles with different size ranges. So this is a treacherous combination. The range of modelled structures corresponding to the series of samples could have particular characteristics changing because of artefacts - i.e. artificial trends produced by the approximations in the software, not relating to your samples. There are a series of important steps necessary to avoid such a pitfall, and it is strongly recommended that you work through them all. The first is to check the quality of input data.