It’s sometimes very difficult to get to the results the same as the ones in a paper. However, I think your results are acceptable and you have a greater Q-factor and better bit error rate than the ones obtained in the paper. I suggest that you go through the paper again and see if there are parameters you have set differently in your implementation. And by the way, the paper’s results have been obtained using Optisystem 9, while yours have been obtained using Optisystem 7. I cannot say that doesn’t affect the results for sure.