You could resample the datasets to have the same resolution and FOV, but this is usually not a good idea because if you resample the anatomical, it'll lose resolution, while if you resample the EPI it will take up enormous gobs of disk space without making your analysis better. The important thing to note is that, essentially, AFNI is storing where these datasets are in space, independent of where the "camera" was when it took a picture. So the pixels don't line up, but the underlying object does. I can't really explain this very well, but the easiest thing to do is open the AFNI viewer and overlay the EPI on the anatomical. You'll see that the anatomical is smooth but the EPI is blocky, because they're on different grids (and you probably have nearest neighbor resampling on because that's the default). But they appear in the same place (you can click "See Overlay" on and off to verify this) because AFNI knows how the voxels in the image map into locations in space.