As the exit message tells you, your constraints are satisfied within the constraint tolerance that you supplied, TolCon=1e-32.
If you don't truly require the bounds to be satisfied exactly, then fmincon was successful and you have your solution. However, using fmincon in this way is massive overkill. You could just as well have used lsqnonlin.
If you do require the bounds to be exactly satisfied, you should use fmincon's interior-point or sqp algorithm with their default settings. Note however, that lb,ub bounds are the only kind of constraints that can be satisfied exactly. Any other constraint can only be guaranteed to within TolCon and, in the case of nonlinear constraints, only when you provide a sufficiently good initial guess, xo.
Also, similar to Torsten comments, it should be unnecessary to have both nonlinear constraints and a least squares cost function that are identical to each other. However, I think it makes more sense to get rid of xcon. If a solution to your equations exists, it should be enough to minimize norm(Research(x)). Since nonlinear constraints can't be satisfied exactly, there is no additional benefit to including xcon. It just adds computational burden.
Best Answer