Identical in structure to Fig. 6, but here showing results for halo mass estimation from the FOF-AM approach. Generally, a good recovery of halo masses is obtained by this approach, but the performance is significantly poorer than with DfL. We also note that similar biases are seen when changing the model - data pairing in FOF-AM to DfL. No uncertainties are shown in this figure since the abundance matching technique does not provide an obvious framework for error propagation when applied to groups (unlike in the DfL method).