I think, as long as the negative values are very low, like in your case, the atom density is in order of e-28, then it could be neglected. You might also see some isotopes have a positive value but within the order of e-200 or e-300 which makes them practically very low to affect your calculations.
Then, the negative value and this low number of atomic densities might came from the numerical solver within the time integrator, that’s why we need to double check our power input and timesteps being used in our depletion calculation.
I hope this discussion could help you