The essential features in generating altimetric sea level time series from crossover differences (as originally proposed by Fu and Chelton, and followed by several other adaptations) are reviewed here. New results are also presented in an effort to answer questions raised in the review. All procedures involve explicit or implicit spatial and temporal smoothing, which are necessary for error reduction and for ensuring reasonably dense temporal sampling rates. One crucial step in the procedures involves the designation of a spatial sampling box, where a larger box results in better error reduction and denser temporal sampling, but may cause excessive signal attenuation. These and various other issues are discussed here by comparing the island tide gauge data to GEOSAT (first 17 months, the nonrepeat era) sea level time series in the tropical Pacific. It is found that the optimal box size varies from island to island because of the geographical dependence of the error as well as the signal. Furthermore the idea of limiting the crossover adjustment process (an orbit error removal procedure) to crossovers between tracks that are less than certain time apart in order to preserve the long‐term, large‐scale oceanic signal is also tested for the first time against the island tide gauge data. The results show improvements around many island stations, albeit stations showing deterioration outnumber those showing improvement, indicating the worsening of the signal‐to‐noise ratio on average either due to lack of signal at the longest and largest scales or due to too much noise at these scales.