The data should have been labelled from the end of each month rather than from the beginning - then there wouldn't be the anomaly on the 31st.

I.e. n[i] should be sum over all days d of tickets[d] where d is the i-th day from the end of the month that d occurs in.