The flow within turbomachines is intrinsically complex and involves boundary layer transition, separation and vortices such as the tip leakage vortex and wakes. In a low-pressure turbine, as the Reynolds number can be small, the flow over the suction side is likely to separate leading to the formation of a laminar (or transitional) separation bubble. This flow mechanism can be predicted using Large-Eddy Simulation. However the computation is still very expensive in a design framework. Thus, Reynolds-Averaged Navier-Stokes (RANS) method is used in the present investigation to simulate the flow over the low-pressure turbine airfoil T106C. The laminar-turbulent transition is modeled with the model of Menter and Langtry. Following the work of Minot et al. in which the CFD setup was deeply investigated, the present study aims at evaluating the sensitivity to uncertainties relative to experimental values (freestream turbulence, Reynolds number, incidence flow angle and exit isentropic Mach number) and at improving this model regarding the calibration of several functions using optimization process. The uncertainty study highlights the parameters which mainly influence the isentropic Mach number and loss distributions. The new calibration of the Menter-Langtry model improves significantly the flow prediction over the suction side, except for the open bubble configuration.