I don't really see why having the energy absorbed in elastic deformation in the C-notch test would be a bad thing. Especially at the higher hardnesses, where there is no ductility to absorb the energy, it can't go anywhere else without fracture.
In the end, you have to use the test that best represents your observations in service. If the U-notch gives better representation of observed behavior, then that's the one to use. The use and review of data from these tests gives you a starting place with a hand full of steels. Further refinement and testing will give a test that predicts behavior for your (anyones) specific use. The best of the narrow choice of alloys in the most representative test is the one to use. Of course it get pretty complicated, since there is almost always more than one criteria, and they typically oppose each other, i.e. grindability in manufacture vs. wear resistance of final part.
Keep in mind that the Charpy family of tests gets a lot of scatter at the hardness levels we like for knives. Maybe not so much with swords, since they typically run at lower hardnesses. The torsional toughness test was developed to represent behavior that was observed at higher hardnesses. The behavior was not predicted reliably by the Charpy family of tests. This follows the same pattern that lead to the Charpy and Izod tests. They were developed to predict behavior that ductility in a tensile test did not predict.