Which top-N list does the performance operator work with for item recommendations?

I am using the "performance" operator of the recommender extension for item recomendation. The values I get for NDCG and MAP are always the same even if I vary the value of "n" in the previous "apply model" operator. My question is: for what value of N of the top-N list does the "performance" operator calculate the value of NDCG and MAP? NDCG and MAP values vary depending on N.