Hello Fayez,
in one case you are running 3E7 primaries in one cycle, while in the other you are running 3E4 primaries in one cycle looping over 1000 cycles.
The total number of particles simulated will be the same. However, during the calculation of the statistical uncertainty for the scored observable, FLUKA takes in account how many cycles have been run. (see also this question in the forum: Statistical error).
Also, consider pages 27-30 from the FLUKA course lecture: https://indico.cern.ch/event/1123370/contributions/4715934/attachments/2444331/4188477/02_Monte_Carlo_Basics_2022_ULB.pdf
Overall, in your specific case, the former run mode would not be suitable for the estimation of the error.
Cheers,
Daniele