But the estimates are based on the 2000 Census. Illegal immigrants may avoid participating in the Census. They likely move more frequently, so that they were no longer around when followup was done. In cases where some members of the household are legal and others not, there might be underreporting. So if a legal resident family has an illegal cousin living with them, he might not show up on the census form.
Non-participation is risky, and participation is not, so I doubt the effect is large. The 2000 census (unlike the 1990 one) was widely applauded as having done a good job of counting everybody, which explains this: