In Witten's paper on QFT and the Jones polynomial, he quantizes the Chern-Simons Lagrangian on $\Sigma\times \mathbb{R}^1$ for two case: (1) $\Sigma$ has no marked points (i.e., no Wilson loops) and (2) $\Sigma$ has marked points and each point has attached a representation of the gauge group. In case (1), Witten shows that the vector space should be the space of holomorphic sections of a determinant line bundle over the moduli space of flat connections. For the second case he states that the vector space should be the $G$-invariant subspace of the tensor product of all the representations associated to the marked points; by this I mean, if $\Sigma$ has $r$ marked points and each point has a rep $R_i$ then the quantum Hilbert space is $(\bigotimes_{i=1}^r R_i)^G$.
Does anyone know how to interpret this second case in terms of sections of some bundle? I mean, shouldn't the second case reduce to the first when you remove the marked points? Also, Witten states, immediately following case (2), that in the presence of no marked points the quantum Hilbert space is 1-dimensional. How can one see that from the formula for the quantum Hilbert space, $(\bigotimes_{i=1}^r R_i)^G$?