The assumption in linear regression models is that the residuals are normally distributed, if everything is working properly. Is it possible to have residuals that are normally distributed when the variables are not? The following example should help clarify the problem.
If you have the equation y=mX+b+e then I want X to be non-normally distributed for any fixed level of y. So if I have fake data, then I might write a for loop for(y in 1:100) to get different levels of y, and the calculate x as uniform for each value of y. Maybe something like x=runif(20,min=1+y, max=5+y). For every level of y, I will get 20 values of x that come from the uniform distribution with a minimum value of 1+y and a maximum value of 5+y. The residuals look sort of normal, but are obviously bounded. I could try another x=runif(20,min=0,max=1)^runif(20,min=0,max=1)/runif(20,min=0,max=1)+y. The residuals look very strange, so this clearly is not a case where the residuals are normally distributed and the independent variable is not.
I suspect that saying that the residuals are normally distributed is equivalent to saying that the independent variables are normally distributed at any level of the dependent variable. This is not the same as saying that the independent variable must be normally distributed.