I’ve already explained stratified random sampling to you on another thread. In this case they are assuming that the August 2012 party affiliation survey by Rasmussen was representative and are doing stratified random sampling based on that weighting of party affiliations, rather than party affiliations from 2008 exit polling data.
The proposal here is not stratified random sampling. They are taking existing polls and simply REPLACING the results with numbers they prefer!
We all encountered this way back in 5th grade when our math teacher ~ the special one brought in from the university ~ referred to "the Egyptian variable" ~ which is the number you multiply your answer with to get the correct answer, or vice versa!