I reckon you'll have to use things like right keep and where not exists (...) to achieve what you want. You'll probably have to use a bunch of resident loads to create new tables with only the data you want.
If you have qvd files available of the tables in the data model of the document you do the binary load from then it may be faster to load those qvd files instead with where not exists clauses so the qvd's will still load optimized.
Thank you for the reply but your solution is not an option for me. With Section Access and the data reduction functionality it works better but I want to get rid of the extra step of saving and opening the reloaded document. I guess it is not possible.
I would expect that a binary load does the same with Section Access as opening the document but it seems not.
Hope someone else has any idea or has a workaround for my issue/challenge.
The reduction using section access depends on a specific user opening the document. Only then is the user known and can be determined which data should be removed/kept.
If you have Publisher available you can use the distribute function to distribute (and thus reduce) by user and achieve pretty much the same result. NPrinting can do it as well I think.