If you really mean "heat" loss, then mismatch is not a source. Mismatch causes some of the energy to be reflected back away from the junction, and thus may be considered a loss to the forward signal path of the system, but it is not converted to heat, and does not, for example, contribute to thermal noise in the system in the same way as dissipative (i.e. heat) loss.
The only other "dissipative" loss that I can think of, besides finite conductivity, is dielectric loss in the medium filling the waveguide. If the waveguide is filled with air, as most of them are, then of course this will be vanishingly small.
Another source of non-dissipative loss could be mode conversion. Note that this occurs even in straight waveguide if the walls have finite conductivity -- not because of the dissipation of the walls, but because it distorts the field geometry near the boundary. This necessarily results in some energy propagating away in a mode other than one you stimulated the part with. If those modes are cutoff, then they too will dissipate in the finite conductivity of the walls.