The easiest way is to use the SUB2NDX function to convert (row,col) subscripts into indices. You didn't specify the dimensions of the final matrix, so in the following example I used a 10-column matrix. You can modify the dimension as you need to:
proc iml;
have = {6,9,1,2};
/* Somewhere you need to specify the columns of want?
Is it max(have)? Is is square? */
want = j(nrow(have),10,0);
rows = T(1:nrow(have));
idx = sub2ndx(dimension(want), rows||have);
want[idx] = 1;
print want;