Hi Rick,
Yes, you are correct. There is just one event per class. Thank you for spelling this out for me---I slapped my head while being reminded that the difference with respect to concatenation in this specific case would be in the t-values, not the betas. This really simplifies and clarifies things, I appreciate it.
And thank you for the run-down on the registration issues. The 3a approach is what I had in mind, just to cut down on an additional transformation step.