My thoughts exactly - about the accuracy. To be honest, 6mm @ 10m is nothing worth considering for metrological purposes. And the point density is also quite limited. In my experience, the accuracy of photogrammetry is quite high if the image set is good. It’s not easy to make a comparison because distance doesn’t really apply here. I would say - without having any concise imperical data - that the accuracy lies within a few pixels, especially in x and y of the images. I once had an error of 1 cm in one of my theodolite aquired ground control points right in the middle of a well covered model of about 10x10 m. I tried everything to make it and fit but RC was battling my efforts up to the point of almost breaking up the component. There was no way I could convince it otherwise - with the closest other GCP maybe 1,5m away. The error remained at about 10mm, whatever I did, even with ridiculously high weight. The mistake was that I did not set the reflector height right. So that proved to me that the internal accuracy is enormously high. Usually, I achieve sub mm accuracy, which is below real world measuring accuracy of the theodolite. Only the margins of a model can be tricky since there are “loose ends”.
Anyway, concerning your troubles with alignment, that’s a bit of a weakness in the learning curve in my view. Because RC wants to be coaxed with the right settings. Exif grouping is one of the most important in my view.