When you run a MCS, it is required to have one equation for each variable that you want to compute. This is a so called exactly determined system of equations In order to be able to perform data reconciliation you would need more equations than variables, which is called an overdetermined system of equations. This could be reached by e.g. measuring/estimating some of the originally unknown variables.
Because you simply transferred the input data of your MCS to STAN, the system is still exactly determined. Thus, after computation, the originally given data should not have changed with respect to their mean value and standard uncertainty. The only thing that happens is that the unknown variables will be computed by mass balances and their uncertainty by error propagation. If you set the unknown values to zero with a large uncertainty, then actually data reconciliation is performed (btw. you are right, this should work). However, you will not notice it because of the large uncertainty these values are the only ones that will be reconciled leading to the same results as before.
Hope this makes things clearer.
