When using EXR mode, both SN and DR settings record in 'M' resolution, except in HR. That's why I do not use EXR mode.
That's why I named this mode 'brain-less'.
apparently, u did not read too much into the implementation behind it...
the hi-iso uses pixel binning.. meaning 2 adjacent pixels are combined as 1... giving it a 6M resolution.
for DR, it takes 2 exposures, and combines them instantaneously to give a 6M resolution...