I think the main reason the values are limited going into the 1.1-1.3 pixel shaders is that the 1.1-1.3 pixel shaders are internally fixed point with something like 10-12 bits so you have very limited range in order to get reasonable fractional precision.

Also, it's still limited on 1.4 too, just not as much so you can pass normals without doing the scale and bias.

With 2.0 and above it's pretty much any value.