Pair-instability (PI) is expected to open a gap in the mass spectrum of black holes (BHs) between ≍40-65 and ≍120 M⊙. The existence of the mass gap is currently being challenged by the detection of GW190521, with a primary component mass of 85+21−14 M⊙. Here, we investigate the main uncertainties on the PI mass gap: the 12C(α, γ)16O reaction rate and the H-rich envelope collapse. With the standard 12C(α, γ)16O rate, the lower edge of the mass gap can be 70 M⊙ if we allow for the collapse of the residual H-rich envelope at metallicity Z ≤ 0.0003. Adopting the uncertainties given by the STARLIB database, for models computed with the 12C(α, γ)16O rate −1σ , we find that the PI mass gap ranges between ≍80 and ≍150 M⊙. Stars with MZAMS > 110 M⊙ may experience a deep dredge-up episode during the core helium-burning phase, that extracts matter from the core enriching the envelope. As a consequence of the He-core mass reduction, a star with MZAMS = 160 M⊙ may avoid the PI and produce a BH of 150 M⊙. In the −2σ case, the PI mass gap ranges from 92 to 110 M⊙. Finally, in models computed with 12C(α, γ)16O −3σ , the mass gap is completely removed by the dredge-up effect. The onset of this dredge-up is particularly sensitive to the assumed model for convection and mixing. The combined effect of H-rich envelope collapse and low 12C(α, γ)16O rate can lead to the formation of BHs with masses consistent with the primary component of GW190521.