Results provided by MILC, MMILC, NP-MILC and NP-MMILC when applied to the maps in Fig. 6. The detection threshold has been set to 4σ_L for all algorithms except for NP-MMILC for which a value of 3.5σ_L has been adopted (see text), and the background has been approximated by a two-dimensional polynomial of degree one. The top and bottom left panels clearly show that in this case both MMILC and NP-MMILC miss only one point source and get rid of the SZ point-like emissions. A greater noise contamination in the detection map of NP-MMILC is also evident. The top and bottom right panels show that MILC and NP-MILC retrieve all point sources but also the SZ point-like emissions. This is a consequence of subtracting the underlying diffuse component with the polynomial approximation. Also for NP-MILC a greater noise contamination of the detection map is evident, and notice the detection of two couples of overlapping point sources in the bottom-right and middle-right part of the map.