The existing methods on the double heterogeneity self-shielding problem adopt the Markovian distribution to describe the spatial distribution of dispersed particles. However, the applicability of Markovian distribution for the practical Fully-Ceramic-Microencapsulated (FCM) fuel still needs to be discussed. This work aims to evaluate the effect caused by the approximation of Markovian distribution on the resonance self-shielding treatment in the situation of filled particles with different sizes in FCM fuel design. To be compared with the Markovian distribution, a series of practical chord length distributions are depicted by the Chord-Length-Sampling method. We implemented the analytical average Dancoff method and the Sanchez-Pomraning Subgroup method for 2D cylindrical FCM fuel. A series of average Dancoff factors and resonance cross-sections are calculated with the reference of results provided by the Monte Carlo method. The results show that the Markovian distribution can be accepted with the normal particle size. In contrast, the Markovian distribution affects the accuracy of resonance calculation severely in terms of extremely large particle size of the dispersed particles.