For those who are interested, this is the connect item submitted to MS

https://connect.microsoft.com/SQLServer/feedback/details/808566/poor-query-plan-generated-for-sql-server-2014-clustered-columnstore-index

Their feedback was to lower the cost threshold for parallelism but this did not work for both the AdventureWorks test case as well as for the real world test case with 69+ million records