How about this?
Create another chunk containing just 'a few' overlapping images from a pair of your existing adjacent large chunks, and align.
Then you should be able to align the two big chunks with the one small chunk using camera based alignment.
Repeat as necessary to get a full set of aligned chunks.
Accuracy should increase with the number of small chunks, and the number of images in each small chunk, but it might be enough for your requirements without doing too much work?
Regarding whether marker based alignment is as accurate as camera based alignment, i guess it depends on the distribution of cameras, and accuracy of your markers, so who knows!