Modern wireless communication systems rely on limited feedback for Fast Link Adaptation (FLA). FLA in presence of advanced receivers needs new physical layer abstractions (PHY abstractions) that are able to capture their performance in order to derive the limited feedback metrics (e.g., Channel Quality Indicators). Recently, a PHY abstraction was proposed for the class of iterative linear minimum mean square error based soft interference cancellation and decoding also referred to as turbo Linear Codeword Interference Cancellation (turbo L-CWIC) receivers in 3GPP. This PHY abstraction needs to be calibrated and a first one-dimensional (1D) calibration method was proposed. In this paper, we revisit the calibration method by proposing a more rigorous framework to minimize the predicted throughput error that dramatically influences the FLA mechanisms. We also suggest a novel multi- dimensional calibration approach. Monte Carlo simulations show that it outperforms the 1D at the expense of the knowledge of the number of independent fading blocks experienced by a codeword.