Most traditional voting methods are preference systems: You take your N candidates and rank them from 1 to N. Every voter is assumed to have a "ground truth" preference of ranking. Tactical voting means that a rational voter may sometimes be encouraged to vote with a ranking other than their true preference.
The weaker criteria that approval voting satisfies is this: A rational voter can always make a rational vote by writing down their true preference ranking in order, drawing a line, and approving everyone above that line. In other words, there is no situation where a rational voter would want to vote A > B when their true preference is B > A. This is a significant bound on the effects of tactical voting.
As you correctly point out, the rationally optimal place to draw that line can be tactical. Nonetheless, there is a strong argument that this is "less tactical" than many other systems allow.