I envisage the setup to be as follows
There is a master copy of all of the files, the trunk. Then you have your branch, where you do your work, this is isolated from the trunk.
When you extract the trunk to a directory, it has all the new changes from other developers (Im assuming you are working in a multi developer environment).
You need to identify all of the files in your working copy of the branch that you have changed, ones that havent changed can be ignored.
The logic in my earlier post was slightly wrong, you may have changed a file that was changed in the trunk earlier by someone else, so simply overwriting it would lose those changes.
So, you need to check out your working copy, merge it in with the trunk (I use the file system to do this instead of repository itself, then commit the changes in the trunk back to the repository).
The merge process is the problem, where you are the only person making changes, this is simple. Im not sure how you manage the conflicted updates, as I said before there are subversion clients that handle this better than tortoise. Here, if the changes are to different parts of the code, it is fairly simple to merge together but when the change has been made to the same lines, I can see no simple way to merge them without doing it by hand.
Depending on what powers you have, you should be able to create a test repository where you can check out the different scenarios.
