WELCOME!
I Agree with DH.
The small, but significant, chemistry changes between formulations and brands will be enough to skew any seemingly meaningful results when swapping at each OCI.
The only way to truly study this topic is to have long term, long duration OCI cycles where multiple OCIs eliminate the variability of brand chemistry. You would have to use both trending and ranges in statistical analysis. You should have an extraction methodology. You'd need blind test data. And the list goes on and on. (I know; I do statistical process and quality control for a living).
UOAs are fun to look at, and can be informational when viewing a single one, but more so as to assure contaminants have not gotten out of control, and that wear metals have not shot through the roof, indicating a major problem. Looking at a single UOA and making a final determination upon lubricant performance simply cannot be done when trying to compare it to another brand/grade.
If you want to convince yourself of just how this problem can manifest itself by single OCI/UOA combinations, pick one fluid and repeat it every odd-numbered cycle, but do the even-numbered ones with different products. You'll quickly find that there's little correlation in results, because the constant chemistry changes will foul the results. (Ex: Mobil, Castrol, Mobil, Shell, Mobil, Amsoil, Mobil, RL, Mobil, Deere, Mobil, Valvoline, etc.) Even your Mobil results won't match up well.
The only way to see how Mobil would perform would be to do a bare minimum of 30 (thirty) successive UOAs all with Mobil, then move on to the next product. See how expensive and long-term this project can become? Figure on each UOA being at approximately 7.5k miles, multiplied by 30 sets of data, it would take 225k miles JUST TO TEST ONE BRAND/GRADE! (Statistically speaking, 30 (thirty) samples is the commonly accepted minimum, to track trends/ranges and reduce input variables to mathematical insignificance). And let's not forget the fact that there are even variations within a brand/grade; even every batch of Delvac 1300 15w-40 lubricant is not the same as the next. The more samples you view of each candidate, the better "true view" you get. The only way to reduce lubricant batch variation would be to buy all 30 loads at one time from one batch. Unfortnately, you can't use them fast enough before most would start to have the add-pack settle out. And by the time you were ready to start the next brand (after 225k miles) it's highly likely that the industry would be on the next iteration of lube level (CK-4 or whatever). Are you starting to see why the big companies can do this, but you can't? They have the resources (time, funds, manpower, labs, etc), and we, as individuals, don't.
Where UOAs really shine is not so much to give a brand/grade indication of "better than ... ", but to track lubricant health and engine health, during long-term use. UOAs used in the correct manner can show you how long your OCI can be sustained, allowing prediction of degradation before a critical limit is reached. Further, you can track wear metals to see what is "normal" for your individual engine. Just like people, each engine is a little bit different due to small manufacturing variations.
I'll provide a somewhat crass analogy, but it's actually fairly accurate: Hopping from brand to brand at each OCI is no different than dating on a series of successive blind dates. It likely might be fun, but there's no real value in it because you learn very little in such a short time period, and what you do learn is likely misleading. True long term relationships take time to develop, and are the only way to reveal what lies underneath.
There are millions of miles of testing on the new CJ-4 fluids by all the major manufacturers, and many entities such as SAE, API, ILSAC, etc. Further, there are large trucking operations that use UOAs for predictive maintenance. Trust me; it's been done far better than you or I could do it as individuals. So, are the CJ-4s "best" for all applications? No. But are they more than adequate, providing substantial performance for the dollar spent, for most applications? Yes, definitely.
My prediction is twofold:
1) you'll enjoy the endeavor, for something to do.
2) you'll not be able to discern which is "better", because your results will be inconclusive (statistically). You might like (or dislike) something you'd see, but you'll not be able to "prove" anything.