I think that the author of the itute solutions is saying that the ray in the 3rd quadrant could rotate within the 3rd quadrant, depending on the position of z/w. So if arg(z/w) is very close to pi/2, then the resultant ray could be very close to the negative y axis.