2 conductor t-line in HFSS (radiation boundry issue)

For some reason, I am having a heck of a time with this:

Its actually a little more complicated than I'll give here, but if you can help me with this, that will be enough.

Ok, simplest thing in the world, two cyliindrical conductors running through a dielectric block. Ground planes on the ends of the block - one of the conductors contacts both of the ground planes. There is a hole in each ground plane - the other conductor is concentric with that hole (so that there is a circular space between the ground plane and 'signal' conductor. Lumped ports are placed such that they are between ends of the 'signal' cylinder and the ground planes.

The problem I'm having is with the radiation boundry. I don't apply it to the faces of the dielectric block where I have the perfE ground planes. I do apply it to the other four faces of the block. When I do this, I end up getting a 14% power loss @ 6 GHz (in a lossless dielectric material, gold conductors).

When I don't set the radiation boundry at all - the power is conserved at all frequencies (the sum the reflected and transmitted power is equal to the power input).

What the heck???????????

If I understand you question correctly when you apply a radiation boundary to the 4 faces of your transmission line some power is lost due to unwanted radiation of the TL. When you don't added any radiation boundaries HFSS automatically assigns ground planes to the 4 faces and the power is contained within the dielectric box.

Yes, YoungEng you're right - the problem is that IDK which of the setups is going to be an accurate representation...

I guess what I'm asking about is how the radiation pattern ends up impacting the simulation...

What sort of boundry condition does it force the field to align to at the faces of the dielectric block? And is that a reasonable assumption?