In GeoPandas, list of latitude and longitude points not being classified into a boundary area

I’m working with the Ceremonial County Boundaries of England shapefile available here: https://www.data.gov.uk/dataset/0fb911e4-ca3a-4553-9136-c4fb069546f9/ceremonial-county-boundaries-of-england

The shapefile is in British National Grid (EPSG:27700).

I also have a DataFrame of locations with latitude/longitude (WGS84). I want to determine which ceremonial county each point falls into.

For most counties, the spatial join works correctly. However, for Cornwall, Greater London, and Rutland, the join always returns NaN, even though when plotted the points are clearly inside the county boundary.

Code

import geopandas as gpd from shapely.geometry import Point regions = gpd.read_file("Ceremonial_County_Boundaries.shp") points = gpd.GeoDataFrame(df, geometry=gpd.points_from_xy(df["lon"], df["lat"]) points_bng = points.to_crs(regions.crs) # Spatial join joined = gpd.sjoin(points_bng, regions, how="left", predicate="within")

Whats going wrong The spatial join correctly matches ~90% of the locations. But all points in Cornwall, Greater London, and Rutland come back as unmatched (index_right = NaN).

I isolated the polygons from the shapefile:

county_poly = regions[regions["NAME"] == "Cornwall"] print("Valid:", county_poly.geometry.is_valid.unique())

Output:

Valid: [False]

And the same happens for Greater London and Rutland — all three return invalid geometries.

Every other county returns Valid: [True] and works correctly.

What i have tried

Confirmed visually that points lie inside the polygon

Plotting shows the points visibly inside the Cornwall/London/Rutland boundaries.

CRS is correct

Shapefile: EPSG:27700

Points: EPSG:4326 → converted to 27700

Any help or thoughts would be appreciated?