I just heard from John Olson, who manages gedmatch. He sent some links to technical publications if anyone is at all interested in the detailed methodology. He also said the following:

The various "oracle" programs differ somewhat in their methodology, but all try to relate your admix results back to the averages of the original populations in the Admixture analysis. Zack Ajmal, author of the HarappaWorld calculator, puts it best in my opinion:

"Do note that the admixture components do not necessarily represent real ancestral populations. ... The best way to look at these admixture results is by comparing individuals and populations."

Note that the Oracle-x calculator is still being developed. The "Option 1" and "Option 2" results use 2 different algorithms for calculating percentages. Eventually, only a single consoldated algorithm will be used (and the results will probably be somewhat different from either of those currently offered).

Jim