Close, repetitive atoms

[Tieqiong]

Description
When calculating PDF using the structures created from ICSD .cif files, there exist some structures that produce PDF with nonphysical high peak at very low r. It was suspected that the structure instances are not constructed as intended, and by investigating using PDFgui there indeed exist multiple instances of the same atom at very close locations. The created structures also have different composition comparing to the .cif file description.

To Reproduce
There are many .cif files from ICSD that can reproduce this issue, for example we can use icsd_236635.cif (Europium Silver Aluminide (1/2.5/2.5)). Using PDFgui we can see the list of generated atoms. Here is what I get using the example file:

.

Expected behavior
We know the composition is (1/2.5/2.5), so the expected list of atom should be something like:
.

Guess
This problem could be caused by errors in parsing error or applying symmetry operations.

Related
Similar but different problem could be found in files such as icsd_059683.cif: