I am not certain what you want.
This approach creates a NaN matrix with the column length of the larger vector, writes the original vectors to it, and then plots them. (I shortened the original vectors because the longer ones took too long (more than 55 seconds) to render here.)
% a1 = randi([0,800],1,2251);
% a2 = randi([0,800],1,2347);
a1 = randi([0,800],1,51);
a2 = randi([0,800],1,47);
maxlen = max(cellfun(@(x)size(x,2), {a1,a2}));
Y1 = NaN(2,maxlen);
Y1(1,1:numel(a1)) = a1;
Y1(2,1:numel(a2)) = a2;
Y1
figure()
X1 = categorical({'red','green'});
X1 = reordercats(X1,{'red','green'});
bar(X1,Y1);
xlabel({'Apples'},'FontWeight','bold');
ylabel({'Number of Apples'},'FontWeight','bold');
I doubt that a loop is necessary.
.
