A diverse array of science goals requires accurate flux calibration of observations with the Atacama Large Millimeter/submillimeter array (ALMA); however, this goal remains challenging due to the stochastic time-variability of the “grid” quasars ALMA uses for calibration. In this work, we use 343.5 GHz (Band 7) ALMA Atacama Compact Array observations of four bright and stable young stellar objects over seven epochs to independently assess the accuracy of the ALMA flux calibration and to refine the relative calibration across epochs. The use of these four extra calibrators allows us to achieve an unprecedented relative ALMA calibration accuracy of ?3%. On the other hand, when the observatory calibrator catalog is not up to date, the Band 7 data calibrated by the ALMA pipeline may have a flux calibration poorer than the nominal 10%, which can be exacerbated by weather-related phase decorrelation when self-calibration of the science target is either not possible or not attempted. We also uncover a relative flux calibration uncertainty between spectral windows of 0.8%, implying that measuring spectral indices within a single ALMA band is likely highly uncertain. We thus recommend various methods for science goals requiring high flux accuracy and robust calibration, in particular, the observation of additional calibrators combined with a relative calibration strategy, and observation of solar system objects for high absolute accuracy.