1. What is the source? Are we assuming an ordinary, well-behaved, impedance-matched source?
2. What is the load? Are we assuming an infinite waveguide, or anything that looks like that (e.g., a very long, lossy (but nonreflective) waveguide, or a matched terminator)?
3. What is the length of the below-cutoff section (in terms of electrical length)?
4. Are we speaking in terms of transient events (in which case, we're talking about the energy of a wavefront launched into the structure, but we're also at the same time necessarily talking about a wideband event), or AC steady state (in which case we should be speaking of power and transmission/reflection only, not energy)?
I'm not entirely sure that length matters, actually. (You probably know better than I do, you've had higher level E&M and more recently than I!
)
If 1 and 2 are true, then the answer is simple: what doesn't "tunnel" through the below-cutoff section, is transmitted; the rest is reflected back at the source. What is transmitted, is just power like any other, but it won't be at the same level as if the source were tied to it directly.
So, obviously, the impedance seen by the source is wrong.
If we insert an antenna tuner (lossless as well), forcing the source to a match, then necessarily the full source power must be transmitted. Can't go anywhere else, it's getting to the output somehow or another. The fields in the front end may be... interesting, depending on how much we've blocked by the sub-cutoff section, and depending on how exactly the coupling element is made, and probably having an interesting frequency response (i.e., we're basically constructing a part of a filter here).
The conclusion should be easy to figure out, given suitably specific setup to the problem; it sounds to me more like a matter of lacking definitions, so perhaps your office should discuss what information is missing, first?
Tim