Using results on soft-collinear factorization for inclusive B-meson decay distributions, a systematic study of the partial B → X s γ decay rate with a cut E γ ≥ E 0 on photon energy is performed. For values of E 0 below about 1.9 GeV, the rate can be calculated without reference to shape functions using a multi-scale operator product expansion (MSOPE). The transition from the shape-function region to the MSOPE region is studied analytically. The resulting prediction for the B → X s γ branching ratio depends on three large scales: m b , √ m b ∆, and ∆ = m b − 2E 0 . Logarithms associated with these scales are resummed at next-to-next-to-leading logarithmic order. While power corrections in Λ QCD /∆ turn out to be small, the sensitivity to the scale ∆ ≈ 1.1 GeV (for E 0 ≈ 1.8 GeV) introduces significant perturbative uncertainties, which so far have been ignored. The new theoretical prediction for the B → X s γ branching ratio with E γ ≥ 1.8 GeV is Br(B → X s γ) = (3.38 +0.31 +0.32 −0.42 −0.30 ) × 10 −4 , where the first error is an estimate of perturbative uncertainties and the second one reflects uncertainties in input parameters. With this cut (89 +6 −7 ± 1)% of all events are contained. The implications of larger theory uncertainties for New Physics searches are briefly explored with the example of the type-II two-Higgs-doublet model, for which the lower bound on the charged-Higgs mass is reduced compared with previous estimates to approximately 200 GeV at 95% confidence level.