Yes, you sort of do understand it correctly. Repeat region summation is only tied to that specific repeat region, which is again tied to the specific model or it's representation, not to the drawing itself really. The actual table is of course inside of a drawing, cause repeat region needs a table to exist, and a table needs a drawing to exist. It's not possible to make repeat regions on model level as far as I know.
If you just want to leverage the repeat region data you already have, then your best bet is to go with third party scripting tool that can read drawing tables using CSV format, then transform the data the way you want, possibly into a one summary table file, and import it back in using the CSV format.
The table coming back from CSV file is gonna look rather messy, so you might need to make a mapkey or two to make it look good again.
One question. Would you not be able to measure the overall mass of all the assemblies from each assembly? Instead of using all the parts one by one. I don't know if you have assemblies filtered out of your repeat region tables, but that could be another way to go.
If you ask me I'd try to avoid all the hassle with repeat regions if possible. I realize the data is already there, and that's why you want to use them, but the data is also in all of the parts used as well. Repeat regions are just slaves for all of these parts. You could make a third party script to open all the parts and read their weight params, and have it summed up for ya, based on whatever conditions you choose. It might be tricky to get all of the quantities correctly, but it can be done, as the data is already there.
The script can be taught to read through as many assemblies/folders as you like.
