The RGB image is looking good Paul - interesting composition in both images. I think your Ha image needs much longer subs (and more of them). I'd recommend a minimum of ten minutes per sub. Also I don't think your field flattner is this issue, but guiding. The ten second guiding correction maybe too long for the mount to guide accurate and deliver pin point stars. Instead of increasing the guider exposure time, try binning the guide chip from 1x1 down to 2x2. Don't be concerned about losing resolution on the guide chip due to the 2x2 binning as software algorithms calculate the star centroid with sub pixel accuracy. Binning will boost the guide chip's sensitivity - critical when guiding through a 6nm Ha filter. You may still need to increase the guider exposure time, but start with five second exposures and work from there. Obviously don't bin the main imaging chip, only the guide chip.
Registar is a remarkable piece of code. I only use it for image registration/alignment purposes, though it does have combine functionality. I use a variety of tools in the processing routine, picking the best fit for the task at hand - unfortunately there is no tool that does it all exceptionally well. An example (a poor one, I should add) of Registar's capabilities in handling different image scales is an image I took of the
Cats Paw Neb. The luminance was acquired using a 12.5" RC operating at ~2800mm FL. I used Registar to align RGB data taken with my FSQ operating at 530mm FL to give the nebula some colour. When you start to get into mosaics, Registar is in a field of its own. Highly recommended.