Embeddable biomarkers are short strands of DNA that can be incorporated into genetic constructs to enable later identification. They are drawn from error correcting codes on the DNA alphabet relative to the Levenshtein metric. This study revisits the Conway variation operator which can serve as a population initializer, mutation operator, or crossover operator depending on its mode of application. The algorithm is applied to a part of the space of code parameters where it had not previously been tested. A parameter setting study establishes that an evolutionary algorithm using this variation operator requires a small population and an intermediate rate of introduction of new material (mutation). Better values, sometimes more than quadrupling previous code sizes, are found for eighteen different code parameters with relatively large word size and high error correction ability. The table of best known code sizes is updated by this study. The parameter study performs comparison with the novel total maximum fitness statistic and a technique for displaying time of last innovation within evolutionary algorithms is introduced.