OK - I hadn't checked which column of entries you had used, but now see you had mixed things up worse than I had thought.
Let's assume we are looking at Australia as an example.
Entrants for DV2015 was 14233, and the derivative rate is 1.45 (because the entrants had derivatives at registration of 6346).
The selected entrants number of 1798 INCLUDES derivatives - which means there were approximately 1240 winning cases (1798/1.45). I'm sure you are going to argue that based on the bolded statement you quoted earlier, but I can tell you I am quite correct.
So - in OC the DV2015 winning chance was 1240/14233 - 8.7%.
However, as you can confirm with anyone with a high OC number in DV2015 or DV2014, being selected is meaningless unless they get to your number. So - the "sensible" chance of selection (a selection that matters) is around 5% or less in OC - because that is all the region can stand in terms of quota - and my statement accounts for the non response rate that you mention. You will notice that the selectees for OC are far lower this year - (in actual fact, too low), but once the entries numbers are published I believe we will see the "usual" number of entries. That means the selectees this year (832) means there were around 573 winning cases, which equates to a winning chance of 4%. At that level, the quota will not be met - again supporting my statement that 5% is what is needed.