There are two different reasons why these population principles state that it might be preferable to add lives of negative welfare. The first, which I referred to as the "Hell Conclusion," is that a principle that values average welfare might consider it good to add lives with negative welfare in a situation where average welfare is negative, because doing so would up the average. The second, which I referred to as the "Sadistic Conclusion," states that, if adding lives with positive welfare can sometimes be bad, then adding a smaller amount of lives with negative welfare might sometimes be less bad.

I am pretty sure I have my terminology straight. I am pretty sure that the "Sadistic Conclusion" the page you linked to is referring to is the second reason, not the first. That being said, your original argument is entirely valid. Adding tormented lives to raise the average is bad, regardless of you refer to it as the "Sadistic Conclusion" or the "Hell Conclusion." I consider it a solid argument against naive and simple formulations of average utilitarianism.

What I refer to as the Sadistic Conclusion differs from the Hell Conclusion in a number of ways, however. Under the Hell Conclusion adding tormented lives is better than adding nobody, providing the tormented lives are slightly less tormented than average. Under the Sadistic Conclusion adding tormented lives is still a very bad thing, it just may be less bad than adding a huge amount of positive lives.

We should definitely reject the Hell Conclusion, but the Sadistic Conclusion seems correct to me. Like I said, people harm themselves all the time in order to avoid having children. All the traditional form of the SC does is concentrate all that harm into one person, instead of spreading it out among a lot of people. It still considers adding negative lives to be a bad thing, just sometimes less bad than adding vast amounts of positive lives.