I think it's phrased a bit weirdly. By sum of the 'difference', I think they just mean they are summing up excess deaths and deaths at each period and then comparing the difference to get the %s. Usually a cumulative graph would be something like this (and I would assume it is here):

Raw data:

Jan: Actual deaths 400; Average deaths 500

Feb: Actual deaths 600; Average deaths 500

Mar: Actual deaths 1000; Average deaths 500

Cumulative deaths as a % of the average cumulative deaths:

Jan - (400/500)-1 = -20%

Feb - (1000/1000)-1 = 0%

Mar - (2000/1500)-1 = 33.3%