The old answer is that the trace identity you give in 2) is not quite right,
Let $R=\sum_i a_i\otimes b_i$ be the $R$ matrix for $U_q(sl_2)$ and let $t$ be the
$4$th root of $q$, then



$$t tr(XY)+ t^{-1}tr(S(X)Y)=\sum_itr(a_iX)tr(b_iY),$$

where $S$ is the antipode.

you can find it in a paper of Bullock, Frohman and Bartoszynska in Communications
in Mathematical Physics in the late 90's where we proved that the space of observables
for lattice gauge field theory based on a fat graph is the Kauffman bracket skein algebra
of the surface which is a regular neighborhood of your graph.  

The skeins are functions on the space of connections on the lattice.  We based our construction of lattice gauge field theory in the work of Alexeev, Schomerus and Grosse.
Also we were inspired somewhat by the work of Buffenoir and Roche on lattice gauge field theory.  In both their approaches, the algebra of observables we defined via generators 
and relations.  They used a Wick ordering to get functions. We saw that the connections were
actually a coalgebra, and the dual product on the observables satisfied the relators given
by the authors in the physical literature. This allowed us to give a coordinate free exposition of lattice gauge field theory based on a quantum group, that led to structural
control over the algebra of observables.  Falling back on an observation of Fock we were
able to show that the algebra of observables quantizes the characters of the underlying surface group with respect to a Poisson structure constructed by Goldman.

To get the signs to work like you want you need to work with $-tr$.  The  minus sign
has been explained nicely by Bonahon and Wong. It comes from the fact that you are looking
at $PSL_2(\mathbb{C})$ representations and lifting them to to $SL_{2}(\mathbb{C})$ representations.

In more modern terms, quantum Teichm\"{u}ller theory as developed by Fock, Checkov, Bonahon
and Kashaev constructs a dual lattice gauge field theory, whose representation theory
has been worked out by Bonahon and his collaborators. What is nice about this is you
can emulate steps of the proof of the geometrization conjecture in the quantum setting
and find fixed representations.  Bonahon and Wong recently proved that the space of
observables contains a large subalgebra which is the Kauffman bracket skein algebra of the
underlying surface.

The shortcoming of all of this is that it doesn't address characters of closed surfaces.
Generally the way you deal with that is by defining a projector.  Ultimately more
geometric approaches via a line bundle over the character variety will probably give more
satisfactory answers in that case.

The quantum hyperbolic invariants of Baseilhac and Bennedetti end up assigning quantum invariants to knots and links in manifolds with a $PSL_2(\mathbb{C})$ representation. When
the underlying representation is trivial these are the invariants of Kashaev which is have been equated with evaluations of the colored Jones polynomial by Murakami and Murakami.