I think current implementation emits diagnostics if the given "module" is not a module, but I'll add some tests to ensure that.
Regarding future directions, here is another thread for that: Using canImport with submodules and specific types.
Why I don't think renaming to canImportModule is a good idea:
If we decide to support checking for availability of types and rename canImport to canImportModule, I would assume we would similarly introduce canImportClass, canImportStruct, canImportEnum, etc., analogous to import class, import struct, import enum, etc..