I believe it uses the interpolation surface TIN that gets created (under cross-sections in Mapper) and converts it to a raster using a sampling technique at the size you specify.
If you interpolate a lot of cross-sections, you have more control over what the terrain looks like as you can add breaklines to the the interpolated cross-sections so it will look better.
The interpolation surface is linearly interpolated between cross-sections, but you can control it a little bit with the edge line layer in Mapper.
It uses the location of the cross-sections to determine everything and I do not believe it uses reach lengths at all.