The anisotropy you speak of from motion does exist. When you extract such a dipole from the data and renormalize though (you can very precisely calculate what the effects of motion would be and fit the direction to the radiation we observe) an asymmetry remains along a different axis.