For the parameters I basically tried to make the DNG as "raw" as possible, so that only the demosaicing would be done by X-Transformer and Lightroom would still handle everything else. If you look at this post I wrote I go into more detail on each setting and why I chose it and have screenshots of my X-Transformer configuration. I went to quite a lot of effort to make sure I understood each setting, so I'm fairly confident in the result.
The built in lens corrections are passed through in the DNG for Lightroom to apply. Unexpectedly though the lens corrections come out slightly differently after going through X-Transformer: Even with "Include as DNG Metadata" set as the lens correction option, when Lightroom applies the corrections from the DNG the result matches what you would get if you got X-Transformer to apply the corrections itself, and doesn't match what you get applying the lens corrections directly from the RAF file. My guess is that when X-Transformer maps the lens corrections from RAF format to DNG format it goes via their algorithm in some way. Although this sounds concerning, in my brief test (again, see the post I linked to above for more detail) the lens corrections applied from the DNG were actually slightly better than when they were applied directly from the RAF.
I'm not overly concerned about deleting the RAF files, as DNG is a more standard format anyway. It's just a shame the DNGs are about 3x the size, but that is unavoidable as they now contain RGB values per pixel rather than the single color value from the X-Trans filter that the RAF stored. Keeping the RAF files is only about 33% more space though, so keeping them isn't a bad option if you have the space.