typically you want to use a projected dataset/data frame when you create a fishnet, but then the x-axis will not be north-south (that is, the longitude will not be static). I've run into this recently myself. My thought is to try to create it using a geographic coord system, then project it. (Have not had the time to try this yet).
Another note, If trying to have a rotated grid, sometimes the rotation can cause the grid to not cover the full extent of you area (e.g if odd shaped polygon). So, when I have rotated, I make my extent larger, rotate, and then for my use, clip by my study area polygon.
if you can supply a picture showing the options you are trying, and a clip of the area you are covering, it may be easier to provide feedback on you input parameters.