It looks like you have to pad with 0 to get 63 full days (you have 62 days + 23 hours), so why not just do that at the beginning, then work with the whole data set as a 3-day array of hour-by-day-by-customer:
ncust = size(x,2);
% pad data with one extra row
x = [x;zeros(1,ncust)];
% reshape data into hour/day/customer
x = reshape(x,24,[],ncust);
% average across days
k1 = squeeze(mean(x,2));
% plot the average for the first five customers
plot(0:23,k1(:,1:5))
% Split data into weekdays (2:6) and weekends (1 & 7)
daynum = mod(1:size(x,2),7);
isweekend = (daynum==1) | (daynum==0);
x_weekday = x(:,~isweekend,:);
x_weekend = x(:,isweekend,:);
At the end, I'm assuming that the days are S/M/T/W/T/F/S. If not, change the 0 and/or 1 in the command isweekend = (daynum==1) | (daynum==0);
