A lot of this has to do with the combination of masks. The problem with your first explanation is that you are visually defining where the border of the "node colour opacity" is. In frame 5, for example, you have a fuzzy zone that is equal to the difference in the max/min altitudes (2,000 & 1,000) you use in the demo, and with a resulting border halfway between two (1,500). At this point the surface has an opacity around 65%. This isn't a bad visual approximation for a single fuzzy zone.
When you then define 2 fuzzy zones, they will interact with each other. In frame 9 of your animated GIF for example, the upper and lower fuzzy zones will have an opacity of 65% at 1,700 and 1,300. But these 2 masks overlap and their combined effect reduces the opacity of the surface so that the "border" appears to be half way, but it's opacity will be less that 65%.
The application of the fuzzy zone to negative altitudes doesn't appear to be quite right. It seems to be behaving as if the fuzzy zone is being shifted downwards rather than starting at the altitude restriction.
In you're example where you use negative numbers for both fuzzy zones, you shouldn't see any surface because one of the masks will always have a value of zero.