Yes, what you have outlined is exactly what my proposal does. There is a need to address this issue. See P1839R0 (I will be submitting a revision in the near future to address some wording issues and to provide more information on the issue)

As for the interpretation of the wording, it pretty unambiguously states that the value is taken in an lvalue-to-rvalue conversion, and it has absolutely nothing to do with how said value is represented.