Objective: Cardiac mapping is an important diagnostic step in cardiac electrophysiology. One of its purposes is to generate a map of the depolarization sequence. This map is constructed in clinical routine either by directly analyzing cardiac electrograms (EGM) recorded invasively or an estimate of these EGMs obtained by a non-invasive technique. Activation maps based on noninvasively estimated EGMs often show artefactual jumps in activation times. To overcome this problem we present a new method to construct the activation maps from reconstructed unipolar EGMs. Methods: On top of the standard estimation of local activation time from unipolar intrinsic deflections, we propose to mutually compare the EGMs in order to estimate the delays in activation for neighboring recording locations. We then describe a workflow to construct a spatially coherent activation map from local activation times and delay estimates in order to create more accurate maps. The method is optimized using simulated data and evaluated on clinical data from 12 different activation sequences. Results: We found that the standard methodology created lines of artificially strong activation time gradient. The proposed workflow enhanced these maps significantly. Conclusion: Estimating delays between neighbors is an interesting option for activation map computation in ECGi.