This is regarding the relative percentage error printed in residual nuclei out put. From the sum.lis, it can be seen that the simulation has been carried out for 2E+8 histories, the error printed in sum.lis file is 0.39%. But in tab.lis file, it can be seen that some radionuclides have more than 10% error. In such scenarios, is it fine to work with those numbers, or should we only choose those radionuclides for which the relative error is below 10%? Or working with even more histories is recommended? Or is there any other way to reduce the residual nuclei error instead of increasing histories?
Some radionuclides in your tab.lis file have larger uncertainties than others, mainly because the processes responsible for their production are less likely to occur than for other radionuclides or their parent elements are less abundant.
To reduce the uncertainty on those radionuclides you can either, as you mentioned, increase the number primaries simulated or implement biasing techniques in your simulation (please have a look at the FLUKA course lecture on biasing: FLUKA Beginner Online Training (31 May 2021 - 11 June 2021): Timetable ยท Indico).
An additional important aspect to consider is also whether you specifically need information on these radionuclides for your application or not. Therefore, if this is the case, and you need an uncertainty less than 10% on a given radionuclide for which at the current stage you have a larger uncertainty, I would recommend to increase the number of primary particles and/or introduce biasing.
Another valid option could be region importance biasing, you might want to have a look at the exercise provided in the FLUKA course where this technique is applied (same link as above) and see whether it can be easily applied to your problem.