Further info:
On further fiddling with test BRIKs that have 3D test objects (like the pyramid with intensity 128) in a volume otherwise filled with low-level (0..20) random noise values ....
It appears the viewer has a sharp change of brightness/contrast behavior at the 1312-or-so bright pixels point. On slices that have a count of bright pixels above that level, the viewer sets the brightness/contrast to something reasonable for viewing.
On slices having less than 1312 bright pixels, the viewer's brightness/cotnrast behavior is suddenly dominated by the background noise. The brighter object of interest (intensity 128) translates to no brighter than the brightest noise (intensity 20) ie: its brightness is clipped .
On such a slice, playing with the manual brightness/contrast buttons never arrives at a setting where the bright object stands out brighter than the noise.
So this suggests some fundamental issue with the viewer's pixel-by-pixel calculation of on-screen brightness, rather than simply poor automatic choice of b/c settings.
Hope that helps.
Graham