Here is the status of the ttbar cross section and uncertainties ATLAS uses:

For the uncertainties:
- cross section is produced with HATHOR, so using Moch et al calculations
- the impact of the top mass uncertainty is done separately - by studying MC produced with different top masses, and then evaluating the effect of an altered cross section and of different kinematics simultaneously. Often, instead of quoting this as a syst. unc., the effect is given for a set of top quark masses instead (so, instead of adding it to the systematics for the PDG value, the cross section is given for different masses around this PDG value).
- PDF uncertainties are 68% confidence intervals for MSTW and CTEQ. I believe only the CTEQ uncertainty is evaluated, and then it is checked that the MSTW value is somewhere within the CTEQ error band)
- scale uncertainties: mu_f and mu_r separately varied between 0.5 and 2 times mtop, constrained to 0.5<mu_f/mu_r<2 ; envelope of variation taken as scale uncertainty
- PDF+scale combined: added linearly
- higher-order corrections: tested but found small wrt other uncertainties: variation of two unknown functions [ Cqq(2), Cgg(2) ], as suggested by Moch

Result e.g. for mtop=172.5 GeV:
sigma=164.57 +4.30-9.27(scale) +7.15-6.51(pdf)

There might be an update on this in the future as there is an "ATLAS SM cross section" task force now...