Reflux (TM) Classifier (RC), one of the advance gravity separator in coal and mineral processing industries, is a combination of liquid fluidized bed, autogenous dense medium and lamella settler. In the previously published literature, the feasibility of the RC for beneficiating low ash Australian coals was analyzed. However, the sensitivity analysis of RC for various operating conditions was not critically assessed for the high ash coking coal fines (0.25-2 mm). Moreover, the clean coal ash and yield relationship is also not established. The parametric evaluation of the effects of important parameters such as feed flow rate, bed setpoint and fluidization water velocity of RC on the concentrate ash, size and yield is investigated in the current work. The separation performance of RC is described through the partition curves, in terms of separation density (rho(50)) and Ecart probable error (E-p). The variation in separation density is also quantified using the slip velocity model. In the laboratory experimentation, the clean coal having ash in the range of 7.9 to 20% was produced from a feed containing around 28% ash content on dry basis. The yield obtained varied from 31 to 64%. It was also observed that the separation density (rho(50)) varied from 1.34 to 1.7 and the Ecart probable error (E-p) varied from 0.07 to 0.1, depending on the operating conditions.