Because an argument isn't the same thing as the data. It's a position on the topic, which by its nature contains within it a selected subset of the data and ignores parts that the arguer either doesn't know or doesn't think important. An argument also has a context, and that context can often be limiting.
Take gay marriage for example. The most common argument put forward against it is that marriage as an institution is centred around procreation. However, as most people now seem to agree, that simply isn't a good argument against gay marriage, because it would otherwise also function as an argument against infertile couples marrying. There are also some other sector-specific arguments (Christian - centric religious arguments, outright homophobic arguments, arguments from tradition) that are similarly weak.
So, how to steelman gay marriage? Construct an argument from a context not usually expressed.
Well, as it turns out, marriage itself has significant negative effects. Beyond the extreme stress of the ceremony itself, the potential for greatly magnified stress during relationship and at its end (divorces are provably more stressful than ordinary breakups of long-term couples) due to a combination of social and legal issues unique to marriage mean a marriage isn't automatically a better deal for same-sex couples.
And as it turns out, the rate of various different mental health disorders provably worsened in the LGBT community in several parts of the US shortly after the introduction of same-sex marriages. So we have a measure which has had a concrete and demonstrable negative on gay people specifically.
Now, is this a convincing argument against gay marriage? Not necessarily. It doesn't confront the human rights issues and it doesn't automatically follow that this will remain the case; it is likely the increases in mental health problems will normalise over time as the additional pressures on the community reduce. But it's certainly a much, much stronger argument against it than virtually any of the usual conservative arguments I've ever encountered. And in the process I've examined my own assumptions, extra bits of data regarding the subject and gone some way towards ceasing to assume opposition to it is driven solely by Bad Evil People being Bad and Evil.
That's how steelmanning works. If your position is solid, you'll still not prove your own position wrong, but the argument against it will be much more solid, and you'll understand the issue better than you did before.