The accuracy with which it is possible to calculate theoretically the positions of energy levels and the low density (in frequency) of the grid on which opacity sampling is calculated both limit the accuracy with which the effect of blends can be determined. For species such as Fe for which a large number of transitions are involved in calculations, this leads to a small effect since, statistically, cancellations occur. However the effect can be large for an element such as Li with only a few transitions contributing to its radiative acceleration, grad(Li), at any given stellar radius. In particular, the Fe lines are shown to modify grad(Li) in and below the Fe convection zone but the modification depends on Fe level positions. This is analyzed quantitatively here using the OPAL data base. The grad(Li) are found to be uncertain by a factor of 2 in Pop I stars but of 1.03 in Pop II stars.