Unless there was some truly horrific coding, it would not be possible for there to be two different sets of tables based on the number of other explorers sent out. It would take having the game store the number of explorers for each person and their recent activity. There would have to be addition variables to control for the time gap between them, as well as player-specific adjustment factors. A normal algorithm for using a lookup table for a random result would essentially one line, with the only variables being the type of search and the random number generated. An algorithm to provide for the tables to adjusted for specific players and for their sending patterns would be at least an order of magnitude more complex.
The answer to the observed data still is most consistent inadequate sample size. This is reinforced by your contrast example, as the other player is obtaining a sampling with at least five times the amount of data and getting percentages that are more consistent with what is expected. This is normal behavior with dealing with a discrete set of returns, and you are still at about a third of size of the sample size needed to look for the change of a single element of the table.