Thanks, Rohit.  In my case, I want to keep the component in the stream.  For example, we have about 20 components and some team members only work on two or three components so they only load two or three components into their stream.  Doing a full load of this workspace only loads those three components (unless in the UI you accept the components or in lscm you use the --flow-components option when accepting).  However, if you load five components and then subsequently change your mind and only want to load three, it will continue to load all five no matter what you do.

what if you browse and select only the components you want loaded? (vs load all)

Sam, I think that would work but in my case the workspace is loaded by Jenkins so the use case is loading the entire repository workspace as it is configured.  It did work to remove the component from the parent stream AND the workspace but to me, this is broken.  A repo workspace should be able to have a subset of components and if you "load all" it will load only the components configured for that workspace.  And it does act that way if the workspace is created initially with the fixed subset of components.  Where it seems to get confused is if the initial workspace is trimmed down of components then it will want to reload everything.  Maybe this is just my world.

Which jenkins RTC plugin are u using?

the open source one loads the STREAM  into the build workspace, and then the build machine from there.  the components are in the stream.

the Product provided plugin gets info from the Build Definition, we we discussed before. you deleted the component from the workspace, and the workspace is connected to the stream.
far as I know, unless you said 'private build' (use ONLY this workspace), the model is load the (flow target) stream into the workspace, and then extract code to build from there.  you might try removing the stream flow target for a test

you did NOT push the component delete to the stream, so you have a content matching problem

Hm, I never noticed that, pretty cool.  However, that's not my problem.  I will do a clean test to confirm, but this is what I see:

I believe Load will load from the stream to the workspace and then into the sandbox.

If I create an empty workspace, and say Load, it gets the components from the stream into the workspace, and then onto my sandbox.  your behavior is consistent with this.

but u said 'build', which loads a(or more than one) STREAM into a workspace,
then the final result of THAT is loaded to the build users sandbox before the build process begins.  again, the two jenkins/RTC itegration tools work differently.
one loads a STREAM (all components), and one follows the Build Definition rules
(all unless excluded, or none if 'private build' is marked)

what happens if you remove the flow target from the workspace before u issue the load?

All four components are loaded.  This is what I expected, only loading the ws configuration. I stopped the daemon and cleaned up (rd /q /s test_workspace).  Then removed one component from the workspace and reloaded.  Hmm.. now it works!  Of course.  Well, via Jenkins it was reloading the ws from the stream.  Guess I can't repro. 

yes, but that is different than the use case u described

" think that would work but in my case the workspace is loaded by Jenkins so the use case is"

jenkins will load from the stream, into the workspace, UNLESS
private build is checked in the connected build def.

please confirm which use case you are working on?  commandline/eclipse
or
jenkins

Oh, I see... that's interesting, it must use the API to load directly from the stream rather than using the workspace?  That does explain the problem.  I "resolved" this by removing the component from the stream which, in this case at least, was suitable.  I will change the question name then I think your answer is right on.  Thank you!