Approach 1 (preferred and should be your default): Credit Value-at-Risk is the distance from the mean to the percentile of the forward distribution, at the desired confidence level (paraphrased from the PRMIA Handbook). This is nothing but the unexpected credit loss at the desired confidence level. The mean in this case is represented by $15m, and the 95th-percentile losses are $40m, therefore Credit Value-at-Risk is also $25m. Your understanding of credit Value-at-Risk should follow this approach as it is the most reasonable and consistent.