That is a really complex question. I'm certainly not an expert in defining the medically relevant genome, but I'll take a shot. Given that we are still in an age of discovery, I'd be hesitant to try to force too much consensus right now. What I do think is important is that test providers and researchers be open about their tests and clearly define what they can and cannot assay with a test. This should include both regions of the genome that may be inaccessible to a test as well as types of variants that may not reliably be called.
This is also why projects like PrecisionFDA can be so valuable. Providing a platform for data sharing and comparison will be useful for increasing our knowledge of what is medically relevant. I would argue that anything that can inform patient care is useful information.