The upshot of the spillover in lordsmerf's and adamdray's LJ's is that some people view the LP expansively in a way that turns System into a combination of the Social Contract and Exploration levels of the Big Model. Social context and implicit influences are part of System. I call this "LP-maximalism." Some people also view SIS as the intersection of all the beliefs held by the players about the game-world, regardless of whether those beliefs have been explicitly communicated. In other words, the term "shared" in SIS is construed as "common".
Against these views, there's the idea that the System only consists of the procedures (Techniques) evidenced in actual game play actions (Ephemera). And the formal definition of SIS is only concerned with elements of the game-world which have been "shared" in the sense of "communicated" or "transmitted" within the context of the game.
Personally, I hold the latter views, and based on the conversation in the Forge thread Shared Imagined Space, Shared Text, as well as the Glossary definitions of System, Ephemera, Techniques, and Shared Imagined Space, I think the weight of Forge authority is on my side (which came as a bit of a surprise).
This isn't to say that social context, psychology of the players, etc., aren't important areas of analysis, nor that the idea of creating a "common" imaginary world among the players is delusional. Just that it's more profitable to separate these social and aesthetic concepts from formal issues of how games are played.
An excellent example can be found in Markus Montola & Mika Loponen's article "A Semiotic View on Diegesis Construction", which can be found in PDF form in Beyond Role and Play. They build on Montola's earlier argument against the objective existence of a single diegesis, or shared ("common") imaginary space among players, in favor of multiple subjective diegeses. In "A Semiotic View on Diegesis Construction", he demonstrates how the subjective diegeses of the various players interact semiotically, often with the goal of making them equifinal, that is, "similar enough to cause indistinguishable consequences". (When I write this, I'm reminded of some of the language in La Ludisto's Interactive Model about "reconcile-and-develop".)
I would also point to works on ritual in RPGs, such as Chris Lehrich's article and posts at the Forge, as well as this thread. These may or may not touch on System, but they're really about the psycho-social and aesthetic elements of roleplaying, including how they interact with System.