When you average, you convert the numbers to pascals. That is a specified part of the standard. It is a bit of a controversial part of the standard though, at least to industry people.
Here is an explanation of the rational for it:
"
Annex A of the CEA 2010 standard recommends adding SPL data in dB to average over the critical bands they refer to as "Ultra Low" (20Hz to 31.5Hz) and "Low" (40 to 63Hz). It is mathematically incorrect to average logarithmic based numbers (such as decibels) as it will bias the outcome to the lowest number in the data set. The correct way to average dB's is to first convert them to Pascals (a linear, not logarithmic, measure of pressure) to properly average the data before converting back to dB's. There is a problem averaging like this, however, since our ears don't hear loudness equally for different frequencies. Averaging in Pascals will bias the average to the highest measurement in the data set. Properly averaging in Pascals (Pa) will make a sub with just one good SPL # still have a good average score. Averaging in dB’s (instead of Pascals) will make a sub with just one bad SPL # look bad..."
In theory CEA-2010 numbers should be comparable, but in practice they are not, at least not to the degree which PSA does on their charts. This is a part of why they are bogus, CEA-2010 is not comparable out to a tenth of a decibel, not even close to that. Its especially misleading when you give your own measurements a 'little boost' and compare that to third party measurements of competitors' products.
If you want to see how CEA-2010 measurements can differ, take a look at different test results for any sub of the same unit. Look at the different data for the VTF15h mk1, that was tested by four different testers. Some of the numbers are similar but some of them are pretty far off. Here is the data in a spreadsheet:
The testers are Hsu, Brent Butterworth (BB), Josh Ricci (JR), and Paul Apollonio (PA). None of these results are wrong, and none of them are 'right'. They are simply what was recorded in those conditions. Now look at the third party data for PSA subwoofers, and compare them to PSA's numbers. I don't have that comparison handy but they are all slightly over Josh Ricci's results whereas Hsu's results are slightly under. Want to see a nice clear illustration of how PSA is gaming numbers, compare Josh Ricci's results of the XV30se and Rythmik FV15HP to the numbers PSA has posted in their
comparison chart.
Btw, I agree that a decibel or two difference in CEA-2010 isn't a big deal, but PSA is presenting the metric as an absolute performance measurement. They might say they aren't, but consider what an uninformed visitor to their website would think.