If a job is significantly risky in RULA, it will be the same for ROSA. If a job is moderate to low risk, the scores from RULA and ROSA could be different. Because ROSA is developed for workers particularly working in a computer workstation, I would go for ROSA.
I have not found any scientific paper which comparing these two methods. In my opinion it’s difficult predict the results because there is many factors which can influence of final results. The Rosa is developed for evaluation of computer workstation and it takes into consideration e.g. the time of sitting, telephoning and so on. Moreover, it is newer than RULA which is more general method. I think if there is not risk at the workstation or very low the results will be probably the same. If risk will be rise the differences between final results of RULA and ROSA will be rise too.
Although, the scores from RULA and ROSA could be different. They would not be poles apart. If you are working with "computer workers" ROSA would be a more appropriate method.