In `division_area`, the `class` attribute might mislead users

Querying for this feature might, under certain circumstances, give users a polygon that includes both land and maritime regions.

See the image below as an example: it's a query for subtype=region for the UK, Belgium, the Netherlands and Denmark.
The downloaded data has class=land, but the shapes clearly include both nautical zones and land cover.

The UK in particular just looks like a blob, and the Netherlands also has a lot of maritime area included by default.

Am I missing something here? Perhaps I am accessing land divisions incorrectly?

Here is the duckdb query (I am using python). Subtype = 'region'

        COPY (
            SELECT
                names.primary AS overture_name,
                class,
                subtype,
                sources,
                id,
                geometry
            FROM
                read_parquet(
                    '{OVERTURE_LINK.format(version=version)}',
                    filename=true,
                    hive_partitioning=true
                )
            WHERE
                country == '{country_a2}'
                AND subtype == '{subtype}'
        )
        TO '{path}'
        WITH (
            FORMAT parquet,
            COMPRESSION zstd
        );
        """

Start with the supplied DuckDB SQL query and the division_area results for subtype=region, especially the class and geometry fields. Compare the UK, Belgium, Netherlands, and Denmark examples to the expected land-division semantics. Done means determining whether maritime areas are included intentionally and clarifying or correcting the misleading classification or documented behavior.