You can, see the next paragraph, "also narrow down around the estimated position iteratively..."

I gave up because I was too lazy to think how to do the "precompute the maximum +/- error bounds" jiggawatts mentions in https://news.ycombinator.com/item?id=34021826