You'll notice that the model I ran did not have your weldment feature at the top of the tube; I used a Total Load At Point (TLAP) to approximate the load transfer through the weldment. I think that, for now, it'd be worth it for you to do this to simplify your model while you work on getting a mesh and model setup for approximating the contact between the tube and the large tube. This will help reduce your element count, thereby decreasing the run-time (even more so if you didn't restrict the p-order on your excluded elements).
You'll find out rather quickly the non-linear analyses take substantially more time to run that linear problems, and that there is a lot of trail-and-error associated with creating a numerically stable and accurate study (hence my suggestion to remove the top weldment components).