From what I've heard, the MS Office applications do not even entirely comply with the published specification. So OOXML is still not a definition of the format with which PowerPoint stores its files.
Your source claims the spec contradicts previous adapted ISO standards, not that it contradicts itself (which is what I assume you mean by it not being self-consistent). It does, however, raise the issue of bits like legacy compatibility for Word 97 behavior and scripts not being documented.
Well, no, it is self-inconsistent in places [0]. However, as you mention, those are probably minor points compared to how it contradicts existing standards, like the Gregorian calendar [1], date representation, and language codes.
Those are annoyances and give increased complexity. They are not, as you described earlier, inconsistencies that make the spec impossible to implement.
Other standards, and web standards in particular, tend to have several open source implementations to guarantee that it's possible to build tools that support it without being the reference implementation. OOXML has no such thing, and the reference implementation is closed source, so it's basically impossible to replicate in any practical sense.
Not fully compatible with PowerPoint, no. That's why I said the PowerPoint format is proprietary; it's impossible to build a working interoperable tool, because either parts of its format remain secret and do not follow the published standard, or the format itself makes references to hidden implementation details.
Having an open source tool that can open files created on PowerPoint without them becoming horribly mangled, or vice versa, is a very realistic concern. Microsoft Office is well known for being extremely difficult to make its file formats portable beyond the very basic layout features.
