This can be made a little more efficient. You're using bisect_left, followed by insort later, which is a combination of bisect followed by insert. So you're basically repeating the bisect step when you could just use the index gotten from the first bisect step. You can also avoid a little bit of confusion by using bisect, instead of bisect_left and remove the +1. The following passes all test cases.

import bisect

def maximumSum(a, m):
    mm,pr=0,0
    a1=[]
    for i in a:
        pr=(pr+i)%m
        mm=max(mm,pr)
        ind=bisect.bisect(a1,pr)
        if(ind<len(a1)):
            mm=max(mm,pr-a1[ind]+m)
        a1.insert(ind, pr)
    return mm