"Optimal" means different things to different people. What the DoT may be trying to maximize (safety, lack of frustration) is generally not what drivers want to maximize (overall speed) and I think there's confusion about that when zipper merge is discussed. The linked article uses the word 'faster' but the Minnesota pdf they cite actually says "Our analysis has shown that the Zipper System has no effect on travel time through the work zone."
Here is a Michigan DoT paper[1] from 2007 that analyzed zipper merge studies in various states and countries. Only 1 out of 6 showed statistically significant increase in throughput.
"It was found that the system did not increase the throughput or decrease the length of the queue." (Netherlands)
"It was found that the throughput did not increase, but the queuing was more efficient and there were less frustrated drivers." (UK)
"Beacher, Fontaine, and Garber (12) found that there was an increase in the throughput and a decrease in the queue length; however, the difference was not statistically significant." (Virginia)
"However, flow did not appear to be significantly affected by the dynamic late lane merge system." (Kansas)
"The data from both of the studies shows that the typical queue length decreased and the lane usage were nearly equal, but the throughput decreased slightly." (Minnesota)
"The results from Maryland’s deployment show an increase in the overall throughput, a reduction of the maximum queue length, and a more even distribution of volume between lanes."
Here is a Michigan DoT paper[1] from 2007 that analyzed zipper merge studies in various states and countries. Only 1 out of 6 showed statistically significant increase in throughput.
"It was found that the system did not increase the throughput or decrease the length of the queue." (Netherlands)
"It was found that the throughput did not increase, but the queuing was more efficient and there were less frustrated drivers." (UK)
"Beacher, Fontaine, and Garber (12) found that there was an increase in the throughput and a decrease in the queue length; however, the difference was not statistically significant." (Virginia)
"However, flow did not appear to be significantly affected by the dynamic late lane merge system." (Kansas)
"The data from both of the studies shows that the typical queue length decreased and the lane usage were nearly equal, but the throughput decreased slightly." (Minnesota)
"The results from Maryland’s deployment show an increase in the overall throughput, a reduction of the maximum queue length, and a more even distribution of volume between lanes."
1: http://www.michigan.gov/documents/mdot/MDOT_Research_Report_...