[vtkusers] vtkDelaunay3D skips points?

[Joerg Rings]

Dear VTK users,

I use VTK for a 3-D tet mesh generator. Here I first create a point cloud and 
triangulate this cloud with vtkDelaunay3D.
My geometry is described with coarse premesh tetrahedra, then the points are 
placed within these tetrahedra, the mesh for each tetrahedron is calculated 
and then collected and merged into a single unstructued grid. This procedure 
has some important advantages over triangulating the whole point cloud in one 
step.
If two premesh tetrahedra touch with one facet, this results in having that 
facet triangulated twice, but the triangulations have to match for collecting 
and merging and should if the input points are the same.
But I discover this not to be true, and it seems like the Delaunay3D 
triangulation skips some points that are in the resultung unstructured grid 
data, but cannot be found as edge vertices of tetrahedra.
It's especially easy when looking at the points of a facet, as the point grid 
is somewhat regular, and I can see points missing.
Now the questions are:
- Does Delaunay3D do such things, so that I can blame it on the filter, or are 
some of the filter options helpful here? I've tinkered with them but without 
any visible change.
- If Delaunay3D is unreliable here, is their a way to create the surface of 
the premesh tets otherwise and initialize Delaunay3D with these triangles? 
It's only important  that the triangles match on the premesh tet surfaces, so 
it wouldn't be that bad if the Delaunay3D isn't perfect inside.

Any suggestion is warmly welcome,
thanks in advance,
Joerg Rings