Convergence Problem: Error smaller than Tolerance

Hi everyone!

I am trying to model non-isothermal, reactive fluid flow in a channel with catalyst-coated walls. Therefor I couple the "Transport of Concentrated Species"-module with the "Non-Isothermal Flow"-module. For solving I chose a fully coupled approach where the linear system is solved iteratively with Vanka-preconditioning.
I set the Tolerance to 1e-3 and the Factor in error estimate to 1. If I try to solve the model, the Convergence Plot of the Newton-iterations shows the error decreasing. The progress of the Nonlinear solver reaches 100% when the error approaches the specified tolerance. However, the progress falls close to zero as soon as the tolerance is reached. The solver does not stop until the maximum number of Newton-Iterations is reached although the error is below the tolerance and is continuing to decrease. When the maximum number of Newton-Iterations is reached I get the error "Maximum number of Newton-Iterations reached. Returned solution is not converged". Can anyone help me with this? If I initialize now with this non-converged solution a segragated solver with the same tolerances the solver converges immediately after the 1st iteration. What is the problem with the fully-coupled solver?

Thanks in advance for your help!

Best regards

Jason