Reducing > 5 million point features to a decent size(fewer points under 50mb) as clusters

Thanks for this awesome library.

As the heading suggests, I am looking for ways to reduce a 907mb, >5 million points file to something under 50mb and fewer points. I have tried different options, but still I get about 350mb , >1 million points file size. I also want the result in clusters.

Things I have tried include:
tippecanoe --output=pv_points.mbtiles -f -Z7 -z15 --drop-fraction-as-needed --cluster-distance=100 -y id -y LCOE ofpv_WASCAL_final2.shp.geojson
tippecanoe --output=pv_points.mbtiles -f --accumulate-attribute=id:comma -Z7 -z15 -B 11 -m 7 -pd --cluster-densest-as-needed -y id -y LCOE ofpv_WASCAL_final2.shp.geojson
tippecanoe --output=pv_points.mbtiles -f --accumulate-attribute=id:comma -zg -B 6 -m 7 -pd --cluster-densest-as-needed -rf1000000 -y id -y LCOE ofpv_WASCAL_final2.shp.geojson
amongst other variants.
I will appreciate if you can point to the best options. I have been on this for over 2 days.

Start by reviewing the three tippecanoe command lines, including their clustering, zoom, attribute, and size-related options, alongside the input file ofpv_WASCAL_final2.shp.geojson. Reproduce the reported output sizes if the data is available and consult the relevant tippecanoe option documentation. Done means identifying and documenting a configuration that produces clustered output under 50 MB with fewer points.