I am going for a trip the next few days so I won't be much online, but to respond shortly of what I mean.
I have given some thought of all these things in the past, the problem of decomposing the observer and the observed, and it sure is true that in standard QM the measurement device (the observer) is a classical system. This is not good enough for serveral reasons.
Nevertheless my opinon is that the problem is not that we have an observer, because there always is one. The problem is howto define the observer. And the solution I envision to this, is that the observer is continiously evolving. The problem of defining the observer, is thus resulting in an evolution.
So with the implicit reference, I mean that any information is relational, and whenever I make a statement, there is always an imlpicit reference to which that statement relates, and this defines an observer.
It's a bit like Zurek's sentiment that "What the observer knows is inseparable from what the observer is".
This means that in my representation, an observer is technically a system of related microstructures, and this system is the observer. So the information is the identifier of the observer. And information is evolving, and identification of observers are a result of spontaneous structure stable formations which are relations to it's environment.
I am still working on this of course, but it at least responds to your question. I am certainly not ignoring this problem, I take it seriously.
/Fredrik