I understand your point. I’m not dogmatic about the physical separation principle, because sometimes it prevents us from mapping useful information about road geometry, but it is still a good rule of thumb.
In the U.S., painted crosshatched dividers often separate the two sides of a two-lane expressway. A chevron divider can also divide the left turn lane from the rest of the lanes; there are reasonable arguments for and against mapping a separate way for the left turn lane. And I’m firmly of the opinion that slip lanes separated by painted pork chop islands should be mapped as separate ways.
Ultimately, the physical separation principle is a good idea, but sometimes we make exceptions because there’s something important we need to communicate on the map about road geometry or movement; legal abstract classifications aren’t the deciding factor but may inform our decision. If the divider is thin enough to function like a standard lane divider and it has a consistent width, then alternatively a tag on a single carriageway can communicate the same information just as effectively.