I recently collected a data set on our KRios G3i equipped with a K3 and GIF. I collected the images in super resolution and saved in lzw tiff format. After importing them with the gain file, the movies have areas of light and dark splotchy-ness. After trying the different varieties of flipping and rotating, it still will not go away, but is common in every image.
The X+Y flip looks correct to me, and it appears the dark applied to the original data and/or the gain was not good. Unfortunately it is probably a permanent feature of every image now. Fortunately, I don’t think for single-particle data this will have much of an effect on things as long as it isn’t affecting motion correction (except maybe particles straddling the seam, but that’s just my opinion).
If you are extremely concerned about it, and you have many images, you can just average everything together (a small bit of outlier rejection might help) and create an adhoc gain. That will certainly make your images look flatter.
If you have a DM .dm3 gain reference, and collect movies with SerialEM, flipping upside down is usually correct (motioncor2 -FlipGain 1). The best tell-tale of the wrong gain alignment are pairs of mirrored, contrast inverted artifacts on either side of the micrograph.
It could be that your gain reference was simply too old. New gains are really needed at least weekly.
I agree with @yoshiokc. If you look at your gain reference directly, is the central line visible? If you are sure your gain was promptly and correctly generated (even illumination at the manufacturer recommended counts, perhaps using lower spot size and C2 excitation), and the line is there in the gain ref itself, you might need a defect file (for motioncor2) or the camera could even require an anneal.