When you simply collect more women then men or an age group is off, then the weight being applied is a simpler correction. When you have multiple categories being off that large of %, then you know the later into the sequential determination of the weights is going to cause a bigger error.
They are claiming an error of +/- 2% when they are required to do post-stratification weighting due to multiple categories off and only have a sampling of 1990 (with some of the questions being shown in the news headlines under 1000 sampling).
See more: Handstee – Trending Shirt