GDH you're right. The essential difficulty is with SWR estimation itself: it collapses decades of data into a single value with,hopefully, some predictive value. The problem is that the historical data has to include not only the variations that you are experiencing with you begin withdrawals...