Some letters aren't letters

[Koen G]

I do lean towards the 'embellished way to represent same sound' explanation. And of course, an embellished P for example might double as a pilcrow.

But I agree that the phoneme inventory becomes rather limited then. If I understand both JKP and Rene correctly, then problematic aspects could be explained somewhat, but all these explanations have the drawback that we'd expect a much larger glyph set if they were true. If that is indeed what is argued, then I fully agree. And anyone defending a natural language hypothesis should take this into account. 

Now there are some ways to explain this. One might be that certain glyphs, like the unusual o's JKP mentions, function like diacritics, modifying the sound of the following glyph. But to be honest I have not studied this enough to know whether it's feasable or not.