Point cluster symbology for different categories without overlapping symbols

I’m trying to visualize crimes based on if the crime happened during day or night time. I’ve tried using the point displacement tool but I’m not satisfied with the results because the readability for certain scale levels is too low and the results also overlap:

So instead I followed Anita Grasers guide and tried using the point clustering-tool but I can’t think of a way to separate day/night crimes without duplicating the layer and setting different offsets:

Ultimately I want to show the clusters including the exact amount of crimes separated by day and/night without having overlapping feature symbology. So I tried to accomplish this for one category by using the below expression but the results are not correct, the total count is not aggregating correctly:

CASE 
WHEN 
"day_night" = "day" THEN @cluster_size
END

Using point cluster symbology in one layer, how can I re-write the expression above to correctly count the numbers of day/nigh crimes separately and visualize it without overlaps?