[Dune] bad grids??

[Marco Cisternino]

Good morning,
I would like to share with you a problem we have.
We have already thought about the solution to it, but it's probably not 
feasible with dune+ALUGrid.
That's why I ask your help, I hope anyone has already had the same 
situation.
We start our finite volume calculation on a coarse grid generated by a 
mesh generator.
We refine the grid in a standard way, dividing fathers. The results is 
that at the border between two different levels of refinement the 
children have the same face slope of the fathers and this means that at 
the leaf level the face slope of these cells change only when the father 
changes.
This kind of grid causes oscillations in dynamical variables with a 
wavelength equal to the fathers grid step.
We've also found out that moving the hanging nodes of the children out 
of the face of the father, using a parametrization, can cure this 
oscillations.
The problem is that we can get elements with 5 faces ( starting with a 
2D grid of quadrilaterals) we are not able to manage with ALUGrid.
I imagine that almost the same situation can be obtained in the 
coarsening of a grid as the result of the elimination of a vertex...
How did you deal with this kind of problem??
Can you cure the oscillation with a discretization argument or the grid 
is the only way???
Thanks a lot for any help.

Best regards,
Marco