I'm working with chromatographic data for the measurement of a chemical that takes two separable forms in equal proportions (a racemic mixture). This produces two adjacent signals (peaks) which are nearly identical, notwithstanding some minor analytical measurement variation. I need to make a judgement call on whether the compound is "detected" or not across many hundreds of samples, so I'm looking specifically at the signal-to-noise ($s/n$) ratios of each peak (using the root-mean-square measurement of baseline noise). I'm using $s/n = 3$ as a threshold for detection (LOD), but some peak pairs close to the threshold of detection will have one peak with s/n barely above and the other barely below.
How can I most appropriately combine these two largely-redundant measurements? Can I average the two s/n measurements for one single easy-to-interpret value?