Proton captures on Mg isotopes play an important role in the Mg-Al cycle active in stellar H-burning regions. In particular, low-energy nuclear resonances in the 25 Mg(p, γ ) 26 Al reaction affect the production of radioactive 26 Al gs as well as the resulting Mg/Al abundance ratio. Reliable estimations of these quantities require precise measurements of the strengths of low-energy resonances. Based on a new experimental study performed at the Laboratory for Underground Nuclear Astrophysics, we provide revised rates of the 25 Mg(p, γ ) 26 Al gs and the 25 Mg(p, γ ) 26 Al m reactions with corresponding uncertainties. In the temperature range 50-150 MK, the new recommended rate of 26 Al m production is up to five times higher than previously assumed. In addition, at T = 100 MK, the revised total reaction rate is a factor of two higher. Note that this is the range of temperature at which the Mg-Al cycle operates in a H-burning zone. The effects of this revision are discussed. Due to the significantly larger 25 Mg(p, γ ) 26 Al m rate, the estimated production of 26 Al gs in H-burning regions is less efficient than previously obtained. As a result, the new rates should imply a smaller contribution from Wolf-Rayet stars to the galactic 26 Al budget. Similarly, we show that the asymptotic giant branch (AGB) extra-mixing scenario does not appear able to explain the most extreme values of 26 Al/ 27 Al, i.e., >10 −2 , found in some O-rich presolar grains. Finally, the substantial increase of the total reaction rate makes the hypothesis of self-pollution by massive AGBs a more robust explanation for the Mg-Al anticorrelation observed in globular-cluster stars.