That's the thing, though. You can't really judge anything by a UOA except if the oil is still good for service.
Patently untrue. I tire of seeing this repeated so very often on BITOG. It is a narrowly focused viewpoint which ignores much else.
There are many reasons to use UOAs.
- track normal wear averages, deviations and trends (micro data analysis)
- compare/contrast individual data points to large sample sets which have established steady averages, deviations and trends (macro data analysis)
- observe for abnormal events (relative to the averages, variation and trends mentioned above)
- monitor contamination sources such as fuel from leaking injectors (or poorly tuned carb), silica from intake issues, coolant from various sources
- monitor the physical condition of the lube such as vis, FP, oxidation, NOACK, etc
- monitor the additives in the lube (those which can be seen by the ICP or other tech) such as Ca, Mg, Ti, etc
Tiny variences in wear metals between different oils are just statistical noise and tell you nothing about how well they are doing their job.
Those "tiny variences" (sic) matter. And while they are not something to get excited over in small quantity, they most certainly have effect in statistical analysis. Further, UOAs have, time and time again, successfully predicted an oncoming catastrophic event; we've seen the UOAs here to prove it. That does not mean they will always detect such an event, but to believe they are incapable of doing so is a near-sighted blunder.
UOAs can be studied in either a macro sense, or a micro sense. (See the "normalcy" article in the UOA subforums). The error which the majority of folks using UOAs commit is ignoring the topic of proper statistical analysis methodologies and limitations.
One UOA sample (or a few, or even several) are NOT enough data to make comparative superlative performance decisions. The inaccuracy of calculating the stdev is absurdly high when small sample sets are in play. Those are what you imply when you speak to the variation quoted above.
It is true that UOAs only see a portion of the wear particles; they cannot see above 5um as a generalization. So they do not see ALL wear. What they see is a sample representation of wear.
Many SAE articles/studies have shown that UOA data indicates good correlation with other methods of tracking wear, such as component weight analysis, electron bombardment, etc. Further, UOAs have shown good correlation to particle count loading.
The main problem with UOAs isn't the technology or the data. It's that folks don't understand the methodology of how to properly use the tool, and what its limitations are.
Don't blame the tool (UOA) for a lack of proper analysis techniques; rather, blame the user. 
