Perhaps it's a matter of working against anticipated modeling order.
Node 10-20... first pipe flange. Anchor on node 10, cnode anchor to 1010. Vertical orientation.
Node 1010-1020... nozzle flange.
Node 1020-1030... nozzle neck. Nozzle flex node 1030, vessel node 1040.
Node 1040-1050... vertical rigid component from vessel head @ nozzle to weld point between head and vessel - which is an arbitrary polar coordinate of theta and r.
Node 1050-1060... rigid element from node 1050 to CL of vessel at head.
It sounds like the order you have in mind is:
1. vessel shell
2. head component, looks like vessel shell
3. rigid element to polar coordinate
4. nozzle neck, and it adopts the previous element's coordinate as the vessel axis.
The current method makes sense, but is somewhat obscure if you don't normally model in that direction, but might not be encountered frequently.
Is that orientation corrected if the nozzle rotates relative to the vessel after the fact? I'll have to test it next change I get.