The nuclear-data uncertainties impact the best-estimate predictions of the nuclear reactor system. In this paper, total uncertainty analyses have been performed for the TMI-1 assembly at both hot zero-power and hot full-power conditions to evaluate the impacts of nuclear-data uncertainties on the predictions of lattice calculations, based on the statistical sampling method. With an improved multigroup cross-section perturbation model, the contributions of various basic cross sections to the uncertainties of k and two-group macroscopic cross sections are obtained. For the total uncertainty analyses, a 172-group cross-section covariance library produced from ENDF/B-VII.1 is used to generate the samples for the multigroup microscopic cross-section library, and DRAGON 5.0 is applied to perform lattice calculations for each sample. The numerical results show that the relative uncertainty of k can reach about 4.7‰ using the vp covariance matrix of 235U-v and 7.1‰ using the vt covariance matrix of 235U-v. The relative uncertainties of two-group macroscopic cross sections vary from about 2.9‰ (for the total cross section of the thermal group) to about 11.9‰ (for the scattering cross section from the fast group to the thermal group). Moreover, through detailed analysis toward uncertainty origins, it has been observed that 235U, 238U, 16O, and 1H are the four most significant contributors, and the uncertainties of 235U-(v, σf, σγ), 238U-(σγ, σ(n,inel), σ(n,elas), v), 16O-(σ(n,elas)), and 1H-(σ(n,elas), σγ) are the most significant cross-section contributors.