Event signatures in IceCube are complex, modulated by both particle physics and properties of the ice and detector. Event reconstruction thus requires accurate modeling of ice properties and detector effects to fit for physics parameters, such as energy and direction. Here, we highlight how improvements in calibration can translate into substantially improving the angular resolution of electromagnetic showers. Since showers are also used to model stochastic energy losses of tracks, we further show how improved ice modeling, along with other track-specific optimizations, leads to more meaningful directional likelihood spaces for high-energy muons. The median angular resolution for showers is improved by a factor of two over an older B-spline model, and accurate directional contours for tracks can be obtained with Wilks’ theorem.