I'll close this issue since it works as designed. If anyone thinks that specifying binary precision is useful, please create a feature request for a `@BinaryPrecision` annotation.