After reading through this thread, I just want to say that it is definitely possible to do research mathematics without a graduate or undergraduate degree. However, so is getting hit by lightning, or being mauled by a cow. The odds are not in your favor, and you will be at a disadvantage compared to someone who did attend graduate school. For example, you will find it harder to get timely feedback from professors, keeping on the right track, or developing a research topic. That's not to say it isn't impossible to do what I suggested, but like I said, anything is possible. Also, while you can learn everything in a graduate program without actually going to graduate school(very hard but not impossible), there is a bit of beaurocracy going on. No one likes to admit the elephant in the room, but success in academia is not completely based on merit but on connections. Without these connections from attending a graduate program or having people back up your research, it will be hard for your paper to get published.
Ultimately, it seems like your main point is that you think it is possible to obtain a graduate education and write a research paper without going to grad school. What you propose sounds good on paper, but we all know that what sounds good on paper may not really in fact be a good solution due to unforeseen technicalities. Also, please keep in mind that in this day and age, there are not a lot of good, quality, research papers published by someone without a graduate education. I agree with you on many points that material in graduate school can be learned if self taught, but empirical evidence suggests that the failure rate for going down such a path is high. Let the buyer beware.