Much of that code will be reimplementation, so the net LOC won't change as much the LOC written. While OSX uses some parts of BSD, which gives it that BSD flavor, it is also substantially a new kernel based on Mach and with many other things under the hood implemented entirely by Apple.
As for the code output, that is on the high-end of what you would expect to see but within the normal range of some highly skilled individual contributors operating within their domain expertise. Of course, this assumes they are largely left alone to do their work and aren't spending half their time in meetings, making powerpoints, etc. Back when I was plying my trade as an individual contributor, I was good for about 8,000 quality LOC per month if I was minimally interrupted, which is only a little less than the engineer in question.
Assuming no lines are removed. I often write tens if lines, commit to trigger a test, then rip out hundreds including the ones I wrote. Netcontribution - negative. I consider that a good day's work.
It seems in 2006 the Darwin XNU package was 1MLOC and the Kext another 1MLOC, so I have hard time believing that 11 others MLOC where needed for the iPhone KERNEL.
Linux 3.13 is 7 MLOC in drivers, 2 MLOC in arch, and only 139 KLOC in kernel, so it's probably the drivers that inflate a kernet size but I doubt the iphone needed so much drivers.
As for the code output, that is on the high-end of what you would expect to see but within the normal range of some highly skilled individual contributors operating within their domain expertise. Of course, this assumes they are largely left alone to do their work and aren't spending half their time in meetings, making powerpoints, etc. Back when I was plying my trade as an individual contributor, I was good for about 8,000 quality LOC per month if I was minimally interrupted, which is only a little less than the engineer in question.