You don't mention this in your experiments, but all Polaroid and Kodak products that use diffusion transfer (except for Bimat) used a set of rails on each side to insure the proper amount of developer or activator is applied to the sandwich. If you don't do this, you chance getting streaks or no image at all.
Land did his initial work using the methods you describe, using conventional camera products but using a prepared reciever sheet. The goo was pH 12 or so in the final product. This agrees more with your last formula example. He used carboxy methyl cellulose in some of his work. Kodak used it in all of their DT products.
The above phenidone analog is not Dimezone-S. Dimezone S has a hydroxy methyl group replacing one of the methyl groups. It is more active, more soluable and more stable.