The MSCI rating is not uniformly distributed in contrast to the S&P and Sustainalytics percentile scores. So I would opine that the linearization you do on the MSCI score forces down correlation/is incorrect. 65% correlation on S&P and Sustainalytics (the only comparables in the sample) is pretty good, considering that ESG is much more heterogenous than credit risk and that non-financial data is poorer quality.