To maximize power transfer for inductively charging electric vehicles, charger and battery coils must be aligned. Wireless sensors can be installed to estimate misalignments; however, existing ranging techniques cannot satisfy the precision requirements of the misalignment estimation. We propose a high precision wireless ranging and misalignment estimation scheme, where high precision is achieved by iteratively measuring, estimating and aligning the coils. Another key aspect is to convert the non-convex misalignment estimation to a more tractable problem with a convex objective. We develop a conditional gradient descent method to solve the problem, which performs gradient descent (or conditional gradient descent on the boundary of the search space) and projects out-of-boundary points back into the space. Employing experimentally validated models, we show that our scheme can achieve 92% the efficiency of perfectly aligned coils in 90% operations, and tolerate correlated distance measurement errors. In contrast, the prior art is susceptible to the correlation, undergoing a significant efficiency degradation of 18.5%.