There is not an easy answer to this question, today, since test methods are still in their infancy, but C 1688 should be the preferred method over C 138.
As C 1688 is still young, and broad testing of pervious concrete is only beginning, there is not a large body of data on field tests of pervious concrete using either C 1688 (or C 138). Further, there is no long-term history as to whether field testing leads to durable pervious concrete pavements. Thus, we don't know if specifying a tight range of voids (+/-3%), at the bottom end of the practical scale (15%), will lead to a successful project or not.
The literature reports that for a pavement to drain, it needs at least 15% voids in the pavement (not a test pot), but doesn't say how to test for 15% voids. So, we don't know if testing for 15% +/- 3% voids in the fresh concrete by C 1688 will lead to a well drained pavement. Additionally, we don't yet have a reliable test method that will give the true void content of a hardened pervious concrete pavement. To demonstrate this, we studied 9 different documented methods for testing voids in pervious concrete pavements on a set of field cores -- the results for void content on a single core ranged from 4.5% to 37.6%. Therefore, 15% voids by one test method, might be 8% or 30% voids by other test methods.
Further, we know that 15% voids by 1688 may lead to 10% voids in the field, or it may lead to 30% voids in the field -- it all depends on how the mix compacts under the contractor's equipment. Thus, specifying such a tight range of voids does not necessarily correlate to a similar tight range in the finished pavement.
Based on all this, I'd suggest following the industry standard for field density (which accounts for voids) as published by ACI 522:
" Fresh density shall be within ±5 lb/ft3 (80 kg/m3) of the specified fresh density. "
The specified fresh density in this case would come from the concrete producer's C 1688 laboratory testing of the proposed mix design.