I can now add some more detailed plots and informations:
The sweep signal for each frequency was performed with 101 points distributed linearly between -12500Hz and + 12500Hz around the central frequency, with a scanning time of 3600s (ie: 36s for each point, with a delta_f of 250Hz)
6 MHz and 8 MHz are visible only in CEB sensor, while 56 MHz is barely visible even in NEB (I and Q only)
WEB does not seem to work at all, NEB seems noisy at 6 and 8 MHz (or does not work), CEB has problems in the frequency _mag (it seems that there is a frequency offset for 6 and 8 MHz, not for 56 MHz) and it also has many artifacts (see a zoom plot) spaced ~100Hz from each injected line
Note: in ALL the plots attached here the color scale is always the same, only to facilitate comparison.