Thanks for the feedback!
The forbidden_elements = "1-*" "hack" should work and thus, this is also partially a bug. Other than that, also a "fix_element_composition" flag or similar might be useful.

Let's implement this feature, as it would also simplify the access to isomers (-> if fix_element_composition is true, we would get isomers only anyway). Brought up by @gasevic.

Points to consider

• Charge should not be set randomly, because for true structural isomers, we don't want IP/EA comparisons but only structural differences
• For the special case, that people want isomers but randomized charges, we can still implement a flag (--random-charges) with which people can turn the random charge assignment on and off and maybe optionally give a preference for closed-shell and open-shell cases (e.g., 0 and 1 (IMHO most default ones))
• We would also get rid of this kinda dirty "trick" that if element_composition is set, AND min_num_atoms and max_num_atoms is in an appropriate range, we take the composition is fixed. This is way to complicated (even though it's now documented better, thanks to @jonathan-schoeps README.md changes).