From the discussion:

Tuning:
 *Agree that this is generator & PDF-dependent
  (and pythia + pythia != pythia + alpgen, but some 
   parameters should be ~stable under such a change)
 *"Rivet+Professor sensitivity test"
   some parameters shouldn't affect certain distributions (and
    don't tune where a generator has known perturbative limitations)
   should we develop a "tuning charter"?

HF:
 *what is the impact of MPI?  (Michelangelo suggestion that TeV Wbb
  "excess" may be due to this)
 *impact of 4 vs 5-flavor scheme
 -> experimental input needed, see below

Exp -> TH:
 *lots of distributions that theorists would like to see (and probably 
  exist, but aren't made public) => theory wishlist
 *preferably these distributions should be unfolded to particle level
  (still better to have them at detector-level than not at all though)
  acceptance and hadronization corrections are optional, but model-
  dependent and therefore not necessarily easy to undo later

TH -> Exp:
 *uncertainties & limitations:
  -scale uncertainties, incl shapes: what is a reasonable central value?
   does taking scale/2 and scale*2 make sense?
  -what about hadronization, merging, ... uncertainties
  -PDFs: even though progress made, not clear issue really solved
   (cfr PDF4LHC recommendation of using all CTEQ+MRST+NNPDF uncertainty
    sets!)  Also only applied at hard scatter level, impact on other 
   aspects?