Thermal diffusivity of GNPs (graphene nano-platelets) is an important thermo-physical property as it is useful to predict the material behavior in many heat transfer applications. GNP samples were pressed at different loads to obtain different densities, and then thermal diffusivity was measured with the flash method. All samples were coated with a thin layer (~1 µm) of colloidal graphite (Aquadag®) on both sides to reduce reflectance of their surfaces and consequently increase the emissivity. Carrying out measurements on both samples with and without coating, a difference between the two series of measurements was found: This is attributed to a non-negligible transmittance of the uncoated samples due to the porosity of GNPs. Furthermore, assuming a spatial distribution of the light within the samples according to the Lambert-Bougert-Beer law, the extinction coefficient of GNP at different densities has been evaluated processing experimental data with a nonlinear least square regression, (NL-LSF, nonlinear least square fitting), whose model contains the extinction coefficient as unknown. The proposed method represents a further improvement of thermal diffusivity data processing, crucial to calculate the extinction coefficient when data with and without coating are available; or to correct biased thermal diffusivity data when the extinction coefficient is already known. Moreover, reflectance effects have been highlighted comparing asymptotic temperature reached during the tests on coated and uncoated samples at different densities. In fact, the decrease of asymptotic temperature of the uncoated samples gives the percentage of the light reflected and consequently an estimate of the reflectance of the GNP surface.