Agree that statement is ambiguous. And I'm not sure of the original article where they did those experiments.
Some interesting & relevant observations on killer strains from one of the original brewery-focused articles from 1973 (!!!)
http://onlinelibrary.wiley.com/doi/10.1002/j.2050-0416.1973.tb03515.x/pdf
1--Killing activity happens fast (within 24h of co-incubation)
2--Killing is influenced by the % inoculated at the start of a co-incubation:
Testing 50/50 and 10/90 mixes of killer/brewery yeasts resulted in complete kill of the brewery yeast within 24h of incubation at 28*C. "Some growth" of the brewery yeast occurred with a mix of 2%/98% killer/brewery yeast; table from the article suggests a mix of ~70%/30% viable killer/brewery yeast at the end of incubation.
In a test fermentation (2L of hopped 1.038 OG malt wort), "marked killer effect resulted only when the inoculum of killer strain was greater than 5%"
Application to us:
Even more strongly supports working hypothesis that CBC-1 is used as a carbonating/conditioning strain. Given the relatively small amount of cells that need to be added, it's probably not nearly enough to kill off all the previous yeast cells that were already present in the fermented beer. Thus the ability to grow/detect S-04, T-58, & WB-06. The fact that growing the dregs in fresh media results in a shift towards CBC-1 predominant samples fits nicely with all this too.