Item 1 is a correct but incomplete answer. Yes you can branch any file from any arbitrary version but the real question is how to you lay down a branch for a whole product based on some previous set of versions. Usually, through either a lable or changelist specification.

Item 2 is correct also but I would like to introduce the two following elaborations: 1) For what you are doing change lists seem to be a better way of specifying the contents of builds. 2) New development should be done off of the "main" branch. The main branch and other reference branches (if there are multiple parallel releases) serve as baselines for releases. The problem with submitting changes directly to a reference branch is that you never know what changes will actually be released. Consider the case where you are nearly ready to release a new version X when the development team is required to release a different set of features Y before X. If you are developing on the main branch then you have problems ... either integrate the pre X versions to main then develop Y or branch Y from the pre X point. It is just much cleaner to develop on a branch and integrate.

Integration with perforce is so much easier than CVS or VSS the extra time is a small price for assurance of a better release.