Removing the quote is not necessarily the accurate way to make the adjustment. The reason is that the blockchain takes the median after the quote has been applied. A witness with a slightly out of date base that is higher (or lower) than average may become the median if their quote is higher (or lower) than average. (Witness price feed updates are not required to be real-time so this is expected.) For this method to be correct you would have to back out all of the witneses quotes and then recompute the median.

The method I described above is correct. I just recalculated it by using the current market price (0.180 as I write this) compared to the latest (final value reported by get_feed_history) aggregate current feed value which is 0.192/1.140. This results in a net discount of 6.5%. As I indicated above this number varies over time but this value is typical. Simply using the current base of 1.140 results in an implied discount of 12.3% which is far from typical.

You are looking at the median of medians, which makes this harder to compute, since there is no true market price against which that is compared. You would have to either take all the feed updates in the past week and back them out, or take hourly true market prices and compute a true non-biased median. The discount range I've suggested has been typical for most if not all of the past week thus the aggregate would likely mirror it.

However, as it turns out the debt ratio is even lower than you point out, since the current market price is higher than the one you used (which is lagged). This was something we paid attention to on the way down, but also matters on the way up. We are about 20% higher than the lagged feed now, so that is a 1-1.5% drop in debt ratio. Good, but still not a great situation.