Avoiding cherry picking with continuous deployment

I'm sure there is a suitable way to deal with this, so just wanted some opinions on the best way.  We currently have a .NET project that has RTC setup with one stream per environment (Dev, QA, UAT, and Prod) using a continuous integration strategy.  Developers are always accepting the latest changes before delivering their code.  As a change is promoted from one env (and stream) to another, the entire code base is compiled into one executable and deployed .  My question is how do I handle leaving out changes that aren't approved, while still elevating changes that are? Ex):

Work item 'A' is put into UAT and approved (app_v1.exe)

Work item 'B' is put into UAT but has failed testing and is reopened (app_v2.exe)

Work item 'C' is in UAT and is approved (app_v3.exe)

How do I promote work item 'A' and 'C' while leaving out 'B' if deployments are done in a 'full code base compile' deployment (meaning app_v3.exe would have changes from work items A, B, and C)?  And how do you account for a scenario where work item 'C' made changes to the same file(s) as work item 'B's?