Left: The performances of the constituent (squares) and area-derivatives (triangles) subtraction methods on the $\tau_{2}/\tau_{1}$ observable, measured with the $k_\mathrm{T}$ reclustering algorithm, are evaluated by embedding PYTHIA MC events into real Pb--Pb data. The filled circles represent the case with no heavy-ion background and the diamonds represent the case with background but no subtraction. Right: The $\tau_{2}/\tau_{1}$ residual distributions characterising the observable resolution due to detector effects and uncorrelated background. Both plots are shown for $40