The continuous-energy Monte Carlo (MC) sensitivity and uncertainty (S/U) analysis conducted using the multigroup covariance matrices has a theoretical pitfall in that it is inconsistent with the principle of continuous-energy MC neutronics calculations because the use of the multigroup covariance matrices means treating covariance data as multigroup variables rather than continuous-energy variables. As a way to get around this deficiency and perform the MC S/U analysis on the theoretically consistent principle, this paper presents a new continuous-energy MC S/U formulation which directly utilizes the continuous-energy covariance data in the evaluated nuclear data libraries instead of the multigroup covariance matrices produced by nuclear data processing codes. The validity of the new MC S/U formulation is examined in terms of the input-nuclear-data-induced k uncertainty of the Godiva critical assembly and the TMI-1 pin cell problem by inputting the continuous-energy covariance data of nuclides involved directly into the continuous-energy MC transport calculations by a Seoul National University MC code, McCARD.