12% in both the 2 person and the 4 person polls are undecided, not voting, or maybe writing in someone - which is equivalent to not voting. The good thing is that there is the same percentage of voters (88%) in both polls.
There is additional difficulty when there are a greater percentage of voters in the 4 person poll than the 2 person poll. (Some of those voters in the 4 person poll are opting out, becoming undecided in the 2 person poll.) Then you have to make an additional assumption as to whether those folks were Stein voters or Johnson voters or a portion of each. And you have to make that additional assumption based on nothing but your gut feeling, so there a good chance of error creeping into the analysis.
If you go to the article itself, Example 2 deals with such a case. You can see how I handled it.
You know you have to be a wonk to even be interested in this. 
Edit history
Recommendations
0 members have recommended this reply (displayed in chronological order):
= new reply since forum marked as read