Does the term "kernel" have many definitions?
I guess that's the core (SCNR) issue here. Kernel is a quite vague definition. It's even visible in aboves picture as Mach itself is defined as a kernel, and everything above (like file systems) services that are added later. Then again, above graphics tie the Mach kernel with services (from the BSD project/kernel) and call the result ('xnu') again a kernel.
So when more utilities get added and bundled as 'Darwin' it is an arbitary coice to call it OS or kernel. In fact, the term 'core OS' coined here (whatever that should mean) sounds much like an attempt to avoid writing again'kernel'.
Bottom line: What 'Kernel' means is an arbitrary decision for a vague concept.
It is one of these 'I know it when I see it' kind of words, concept words, that are great for a generalized abstract discussion, but do not work for an in detail description.
So for a concrete product, it's whatever parts the manufacturer calls so. And Jobs as CEO does hold some stake here, doesn't he?
(And as John Dallman already mentioned a term for simplification and marketing, bendable as needed, to sell the product with some techno babel aura)