Thanks, Arguile! Perin's article was right on track, and helps me rephrase my question. Perin states:

The pipeline style does all the work up front in a standard CGI or mod
+_perl handler, then decides which template to run and passes some dat
+a to it. The template has no control flow logic in it, just presentat
+ion logic, e.g. show this graphic if this item is on sale. Popular sy
+stems supporting this approach include HTML::Template and Template To
+olkit.

The callback model works well for publishing-oriented sites where the 
+pages are essentially mix and match sets of articles and lists. Ideal
+ly, a site can be broken down into visual ``components'' or pieces of
+ pages that are general enough for an HTML coder to recombine them in
+to entirely new kinds of pages without any help from a programmer.

[snip]

The pipeline model is more like a traditional model-view-controller de
+sign. Working this way can provide additional performance tuning oppo
+rtunities over an approach where you don't know what data will be nee
+ded at the beginning of the request. You can aggregate database queri
+es, make smarter choices about caching, etc. It can also promote a cl
+eaner separation of application logic and presentation. However, this
+ approach takes longer to get started with since it's a bigger concep
+tual hurdle and always involves at least two files: one for the Perl 
+code and one for the template.
[download]

So my question is:
am I on the wrong track to consider Mason when I want to use a pipeline model, eg M-V-C?