I'm trying to understand a bit more on the modeling of a tied universal expansion joint in CAESAR II. I'm modeling this per the CAESAR II User's Guide, quick description as follows:

1) Modeled along X axis for reference
2) Modeled in order flange (rigid), bellows, centerspool (pipe), bellows, flange (rigid)
3) Weightless rigid through entire joint from flange to flange, rigid is connected to one end and c-noded to other end with X, RZ, & RY restraints c-noded to the other flange
4) I've input the "individual bellows" stiffnesses into each bellows, leaving the Bending Stiffness blank. Theses stiffnesses were obtained from EJ vendor I'm working with, specifically these are the individual bellows (not "total joint" stiffnesses).

We're modeling 2 of these type of joints on a 650F @ 30" NPS line. For simplicity, imagine an L shape with an anchor on each end, with one joint in each leg to take up the thermal growth of the perpendicular legs.

Questions:
1) I do notice a difference (~10-20% increase) in anchor loads if I go back through and enter the bending stiffnesses that were provided by the EJ vendor. However, I do get warning 66E when i fully define all stiffnesses. My preference is to enter all of the values since I have them vs letting CAESAR calculate the blank one. Is one approach more accurate than the other?

2) I'd understood the lateral flexibility from a tied universal joint comes from the ability of the centerspool to angulate relative to the bellows. I've tested this in CAESAR by doubling the length of the centerspool (and the associated overall joint length, weightless rigid, etc) and did see a reduction in the anchor loads. Is there anything I'm missing in terms of how the CAESAR software approaches this (does this modeling also angulate the center spool)?

3) Just curious, the EJ vendor did submit the "total joint" stiffnesses as well. I did a test run where I'd modeled a single bellows, same length and tie-rod modeling as the tied universal, but input the "total joint" stiffnesses into the single element. I found that the overall anchor loads increased significantly. I'm thinking I should stick with my more detailed modeling of the tied universal joint, but just curious if anyone else has thoughts on this approach.