These pressure balanced units are not very common. They are complex and every time when I design them I have to "reinvent them". I try again :-)

Sam's post above has an image. In this design one set of nuts is away from the plates. They should be at the plate. The idea is that when the unit is compressed or expanded volume stays the same e.g. 2 small bellows units compress and the big unit expands. Small units have lower spring rate. If the nuts are loose and there is no pressure then small elements compress and nothing happens to the centre bellows. See http://www.jat.co.za/worksamples.htm for similar design. (Internal forces on rods and plate can be huge.)

If you position the EJ vertical, no pressure inside and top flange is loose then it will get shorter due to its own mass. Amount can be calculated using spring rates. As a result if the top flange is bolted to something it will pull that part downwards. For me by putting support under vertical EJ I would use 1/2 mass on top flange as downwards force.

Imbalance under vacuum or internal pressure will add to end forces.

Forget internal friction forces in vertical installation.

You need to add spring force e.g. spring rate * axial movement. Manufacturer has to give spring rate. They can calculate EJMA spring rate, which is not working rate. Working rate is less. It cannot be calculated, only measured. Assume 30% tolerance on given spring rates.