Baryon acoustic oscillations (BAOs) in the early Universe are predicted to leave an as yet undetected signature on the relative clustering of total mass versus luminous matter. This signature, a modulation of the relative large-scale clustering of baryons and dark matter, offers a new angle to compare the large-scale distribution of light versus mass. A detection of this effect would provide an important confirmation of the standard cosmological paradigm and constrain alternatives to dark matter as well as non-standard fluctuations such as compensated isocurvature perturbations (CIPs). The first attempt to measure this effect in the SDSS-III BOSS Data Release 10 CMASS sample remained inconclusive but allowed to develop a method, which we detail here and use to conduct the second observational search. When using the same model as in our previous study and including CIPs in the model, the DR12 data are consistent with a null-detection, a result in tension with the strong evidence previously measured with the DR10 data. This tension remains when we use a more realistic model taking into account our knowledge of the survey flux limit, as the data then privilege a zero effect. In the absence of CIPs, we obtain a null detection consistent with both the absence of the effect and amplitude predicted in previous theoretical studies. This shows the necessity of more accurate data in order to prove or disprove the theoretical predictions.