Hello,
Your current workflow is the one we have used in the past and is the one we continue to use. One point in your workflow which does differ is the selection of your vertical coordinate. If you are using a VRS network through OS-Net you will be receiving ellipsoidal heights referenced to GRS80 – the ellipsoid used with ETRS89. When using OSGM15 you are transforming from GRS80 to Ordnance Datum Newlyn (ODN heights), not to Airy 1830 ellipsoidal heights which is the ellipsoid used with British National Grid. Hence you should select arbitrary as your vertical reference.
The Pix4D transformation from WGS84 to OSGB36 does not use the OSTN15 and OSGM15 transformation models, meaning if you were to import ETRS89 ellipsoidal RTK data from the exif/xmp and select your output at OSGB36 using no ground control, your results would not be as accurate as using the OSTN15 and OSGM15 transformation models. As far as I am aware, Pix4D uses a Helmert 7 parameter model.
If you are working in a local grid it is likely there will be defined transformation parameters between ETRS89 and your local grid. I suggest that either you convert your images to this local grid, or ensure your GCPs are in the local grid and as you rightly say, use those to bring the whole survey into the local grid.
I would always recommend using ground control regardless of your coordinate system, especially if it is a local grid. Without checkpoints you will not be able to say with any confidence what your estimated project accuracies are.
For more information, please feel free to contact me at jason@geodronesurvey.com