So, the root problem lies in the get_candidates_vectorised function. The rapidfuzz library actually returns output based on the case-sensitiveness. So you need to change this function to ensure entire central is not filtered to elimination. (Add .lower() to each bank_make and x)

def get_candidates_vectorized(bank_make, central_df, threshold=60):
    # Use fuzzy matching on make names
    make_scores = central_df['make_name'].apply(
        lambda x: fuzz.token_set_ratio(bank_make.lower(), x.lower())
    )
    return central_df[make_scores > threshold].index.tolist()