This is what DxOMark is showing about their effective ISO measurements:
That behaviour (a high effective ISO, assumed to correspond to an equally higher than normal ISO gain), was typical on old sensors to maximize captured DR at risk of clipping some highlights.
But when we look at equally exposed RAW files on the 1DX III and 1DX II, they are nearly identical (same highlight headroom and SNR performance):
Something seems to have gone wrong with DxOMark testing on the 1DX III.
Regards