A few things to think about:
1) Adding a tilt corrector between the corrector and the camera is only useful for address a sensor that is tilted. It will not help you to resolve tilt issues that exist due to focuser or OTA collimation error. Have you ruled out whether the sensor is the issue or not? You can do this by rotating the camera 180 degrees and leaving everything else the same. Does the tilt follow the camera or does it remain where it was?
2) How are you collimating your scope? Are you using a laser and/or a cheshire? Newtonian collimation is tricky in the first place. Try to add a reducer and image at f2.8 and that amplifies everything. A laser/Chechire is not accurate enough to collimate a Newt for imaging at this focal ratio. If you are using a catseye collimation set with hotspot on the primary, Autocollimator with offset pupil and Cheshire... you have a prayer of collimating the scope to be able to image at f2.8. Collimation of the OTA is the number one issue when people experience issues like this. It is less likely to be tilt in the focuser, but that is always a possibility. Collimation must be perfect to rule out whether there is any collimation error due to the focuser. Re-read that last sentence. Then re-read the sentence that I bolded.
3) If you are able to achieve perfect collimation and you still have tilt you can start to investigate the focuser. With a moonlite focuser (or any that allow collimation of the drawtube and tilt of the base) you will be able to address issues that reside in the focuser. There are two types of collimation error in a focuser . Tilt of the entire body, as well as axial alignment of the drawtube. If the drawtube is not collimated accurately, you can cancel that out by tilting the base, but you will be unable to rotate the focuser to maintain collimation. Collimating a focuser drawtube is tricky, but with a high quality laser and by removing the primary cell you can shoot a wall target off the secondary mirror and dial this in.
The main thing here, is that most issues truly lie within the the collimation of the OTA. If you are not using a system that amplifies collimation error, such as the catseye system, there is no way you are going to get perfect collimation with traditional tools and be able to image at f2.8... unless maybe you are Vic Menard. That said, can you guess what tools Vic uses for collimation? And as far as I know, he does not try to image at f2.8.
Lastly, the reducer you are using is not designed to produce pin-point stars across the frame. It's just not up to the task. Whether you have the ASA version, or the TS version, odds are if you correct all the error I discussed above, you will not get perfect stars. If you want speed, you will have to deal with less than perfection. If you want perfect stars, get a paracorr and image at f4.5. If you want very good stars, get a GPU/Quattro and image at f4.
All of this is stated, without seeing your issues in an image. These are just my thoughts having imaged with three different Newts, three different focusers, 4 different correctors (including the reducer). Good luck.