Modelling expansion joint

While modelling expansion joint following caesar Ii application guide I find that in the tie rod and flage connection translational restraints are used. But if I am modelling using the method described (page attached) the expansion joint is not working as an expansion joint and failing in expansion stress (caesar file attached). On the other hand if i use caesar module even with a single tied expansion joint the expansion stresses are within allowable limits. Then I changed the translational restraints of tie rods into rotational and in this case the expansion stresses are within allowable. So which procedure is actual and to be followed. Pls indicate if i am making any mistake in first modelling.

For your application, at node no. 3500, 3510, 3520 and 3530 you should have X Y Z restraints. A full anchor will prevent the rod to rotate therefore will not work laterally.

Replacing translational restraints with rotational restraints is not in accordance with reality, the rods are guided therefore have translational restraints.

Regards,

see this:

http://www.flexatherm.com/images/universal.swf

But in Caesar II application guide no restraints are used for those nodes..Why?