well, as you have no doubt figured out, you can't really calculate a standard deviation for the variation from a DoE as your data is VERY chunky; you will have a few points at only a few levels of the input range.
If I were forced to calculate an index, I would determine the minimum and maximum output value that I would expect if I set the inputs at their levels that guranteed in spec results. Then I would calculate a simple ratio of the tolerance range to the max-min output range.