Essentially that filter is probably unusable. Let’s do some math to help us both realize that who ever wrote the filter was probably running on a VERY large memory machine.
Let us assume that you have a microstructure that is 1 micron on a side and you are using a lattice constant of 1 angstrom. That means that there are at least 10000 x 10000 x 10000 atoms in that small microstructure. This is 1x10E12 atoms or 1 Tera worth of atoms. Since we store the location of each atom as 3x4byte floats, that is 12 bytes for each position which gives us 12 Terabytes of memory that is needed just to allocate for this filter.
Unfortunately the InsertAtoms filter was hard coded to assume the position values that were passed into the filter are in micrometers so there isn’t much you can do. If your microstructure is really in nanometers you could try to save the .dream3d file, use something like Python to open the .dream3d file, find the vertices of the mesh and scale them to be expressed as nanometers instead. (So multiply each vertex by 1000). This might get you close as you would remove 9 orders of magnitude from the calculation.