yes the reason is just more precision. Ideally you would use a floating point texture, but right now I dont know of an easy way to generate them.

For instance, most topographic raster data is stored in various floating point or 32-bit formats...most DEMs are in such formats, from like the USGS and the space shuttle topography data..(this is cool stuff btw).

Also, progs like World Machine and Terragen use a floating point format to store heightmap data. There is however no easy way to convert these formats into a game usable 32-bit or 24 bit image. The best I can do is try various height gradients, such 0,0,0 -> 0,0,255 -> 0,255,255 -> 255,255,255 so that you have a full use of the color channels. However my attempts with this have not been all that improved form the 8 bit, and sometimes overly "chunky" or rounded.

Also, think about this: normal maps derived from simple 8 bit greyscale have alow precision as well.. you can make better normal maps using rgb channels..(if you want to make the normal map form a heightmap).