Hi Marco, it is actually a lot more difficult to calculate the shell on such things that you might think at first. MoI's sheller is not very advanced, it just does not have the facilities inside of it to calculate an offset where the result has a different vertex structure than the starting shape.

Here is a bit of an explanation for what I mean. Say you take just these 5 separated surfaces:



Then if I offset them by 10 units as individual surfaces, I will get this:



Here is another view:



Note how there is not an equal amount of spacing in the gaps between each piece - that means that these pieces do not meet up in a single point even though the original surfaces met at a single vertex point.

If I untrim all of these surfaces to get their underlying planes, and look at it from the underside, you can see this:



So this is the kind of structure that would need to be created in this case:



I will take a look at it and see if I am missing something in using the shell function that would handle this kind of "deviating vertex" type situation, but I think that you may need to use a different program that has a more advanced shelling function than MoI for doing this.

If you offset by only a very small amount, then those 2 deviating vertices that I showed above do not get separated by very much distance, and if it is not too big of a distance that is not too much of an error to have them represented by just one vertex, that is why using a smaller distance can produce an ok result, although probably the one that you had problems doing STL with had too big of a gap between some of the trim boundaries and the vertex.

- Michael