[Evolve][Doc] Coordinated omission data description

Signed-off-by: Begley Brothers Inc <begleybrothers@gmail.com>

README.md

@@ -610,7 +610,11 @@ Use a [Kolmogorov-Smirnov](https://www.itl.nist.gov/div898/handbook/eda/section3
 
 The KS and AD tests assume the two data samples are independent of one another. However, Goose produces the CO-adjusted data from the raw data.  Hence, obviously, the CO-adjusted data is not independent of the raw data produced in the same test/run.
 
-Goose also shows multiple percentile graphs, again showing first the "raw" metrics followed by the "adjusted" metrics. The "raw" graph would suggest that less than 1% of the requests for the `GET (Anon) node page` were slow, and less than 0.1% of the requests for the `GET (Auth) node page` were slow. However, through Coordinated Omission Mitigation we can see that statistically this would have actually affected all requests, and for authenticated users the impact is visible on >25% of the requests.
+There are situations where absolute values of a percentile are of interest, e.g.service level agreements, irrespective of the circumstances. Consequently, Goose produces percentile tables, showing the "raw" metrics followed by the "adjusted" metrics.
+
+Returning to the example data. The "raw" graph indicates less than 1% of the responses to requests for the `GET (Anon) node page` were as slow as 3 seconds or worse, and less than 0.1% of the responses to requests for the `GET (Auth) node page` were as slow as 3 seconds or worse.
+
+However, the data generated by Coordinated Omission mitigation indicates **2% of responses** to requests across all pages **were delayed by 2 seconds or worse**. For authenticated users **>25% of responses** to requests were **more than ten times slower than the raw data indicated** (comment form posting being slightly less affected).
 
 ```
  ------------------------------------------------------------------------------