To put your problem in general terms, suppose you have a row vector of values, v, with some given sum. For the i-th element you wish to change from v(i) = vi1 to v(i) = vi2 and adjust the remaining elements proportionally so as to retain the same overall sum. Then do this:
s = sum(v);
f = (s-vi2)/(s-vi1);
v = [f*v(1:i-1),vi2,f*v(i+1:end)];
Then v should have the same sum as before (except for possible tiny rounding error differences.) (We assume that the sum is different from vi1.)
