I'm inclined to think that that wouldn't work. The flywheel gaining acceleration on one side of the gravitic gradient would be counteracted by the flywheel losing acceleration on the opposite side. Just like positive normal mass on one side of the flywheel doesn't give free energy, neither would negative effective mass.
The Maxwell equations are known to be approximations, but it was thought that the errors were too small to matter in real life. However if those errors were able to be compounded somehow, interesting things might happen. As far as I gather, the secret behind the EM Drive is that strong (oscillating? shaped?) magnetic fields are capable of warping space and can do it enough for it to be noticeable and practical.