That's what I figured you had implied earlier. Hence my comment, "The assumption that all 20K are going towards the backlog isn't accurate given we know CIC processes applications in parallel i.e. backlogged mixed in with new."
Since we know that applications are being processed in parallel, it is reasonable to see not all 20k are going to reduce the backlog. Only portion of that 20k are being used to reduce the backlog. So, let's say 10k of those 20k (for argument's sake) are being used to reduce the backlog, then 15% reduction by those 10k slots would mean a backlog of under ~70k.
Again, my point is that we know CIC will use that 20k for both new and old applications. Using past numbers, even then the backlog was never this big. 15% reduction is because of a number smaller than 20k.
Just offering my contrarian opinion ...