Thanks for that matt, your input is always helpful and thoughtful - Sorry to strain your memory so unfairly.
I built a study model to try to get to the bottom of why side tangency is not available to me in this situation. Turns out that GC side tangency for surface loft DOES work in my 2003 study model (built entirely from surfaces). The real model is one in which I've had to harvest surfaces from the face of a solid model built by the client, and I thought perhaps Side Tangency didn't work in mixed models, but when I added a solid to my study model, it didn't break the loft.
Actually 2003 does have Fill, and intersection curve, so I didn't have to resort to the right-angle turn (in any case, I wanted end control as well). There were only a couple of faces of the loft where the lack of side tangency was causing problems (inability to do a thin shell) so I deleted those faces and patched them with Fill and was away again. I do hate having to complicate the model with these arcane workarounds, particularly as my client has to run with the model as part of an assembly, but, hey, life would be boring if everything could always be done elegantly.
The only thing in later versions I really miss for this project (fairly complex sheetmetal deep-draw pressings, mixture of prismatic and flowing) is variable radius fillets for surfaces. I'm having to drop back into solids at various stages of the model for this reason alone, but it's a real pain that variable radius fillets also don't work for multibodies - I end up having to delete and recreate the same solid over and over to work around this.