[swamp2]

Welcome back RG. I've taken a long break from the forum myself as well. There was some crazy stuff happening with moderation (IMHO) and the whole S55 under rating debate.

Anyway, thanks for the additional data. More data always helps. However, some criticisms as well:

1. 100% of your observed journal sizes fall within 3σ, there is not 3% of the data outside of 3σ, period. As with any normal distribution, by definition, about 0.2% of all data (both high and low sides will be outside of +/- 3σ. This is far less than one observation out of about 100 and thus no observations are outside of 3σ.
2. Tolerance stacking is a completely universal phenomena for every dimension of every assembly in every industry. The mere observation of and even quantification of the amount of tolerance stacking present simply IS NOT evidence that tolerance stacking is the cause of failures. The reason for this should be obvious. There is no yardstick/metric (that we have) by which to measure the effect of tolerance stacking. A metric would be something like an observation correlating engine lifetime vs. measured bearing clearance, where clearances are known before the engine is put into use. No one has ANY sort of data along these lines for the S65. Yes we can surely say that with no clearances (or perhaps just enough to allow rotation) an engine will fail almost immediately. We also know that with somewhere in the range of .001 in/in clearance (in of clearance per in of journal size) can result in an extremely long engine life, say perhaps 300k-500k mi or even more. But in this case we have nothing in between that identifies the strength of the effect or an obvious cut off point of the effect. So for this reason along with point #1 above, your 3% of engines failing based on being 3σ "out" claim is both incorrect and is not justifiable. Furthermore, even if your observation about the statistics were correct, the conclusion does not follow. I've also not seen any evidence whatsoever that the to date failure rate is anywhere close to 3%. It has not been updated in some time, but back in early 2014, I made the most careful calculation I've seen, taking into account forum ownership rates, actual empirical observed failures and other factors to conclude a failure rate of about 0.1% - 0.5%. Link to that post here.
3. In short we are left with very little additional proof of cause. There is nothing here that moves this argument beyond speculation. The updated *nominal* journal sizes here result in a 0.83 thousandths of an in/in which still resides within/above the prior identified minimum Clevite spec of 0.75 thousandths of an in/in. And certainly it does not make sense to "cherry pick" the tight end result of the spread in clearances to say the S65 clearance is less than the minimum Clevite spec (the tight end is in fact less than the Clevite minimum). Why can't we do so? That spec is based on NOMINAL DIMENSIONS, with the obvious understanding that tolerance stacking always exists! Some additional data that would be great would be an empirical observation that the standard deviation of BMWs journal sizes are significantly larger than an industry norm (and I highly doubt they are, my guess is that they are smaller).
4. I'm still of the belief that BMW knew exactly what clearance they wanted, fully understood the tradeoffs with engine longevity vs. range of clearances they would obtain from their and their suppliers individual part tolerances and then chose the values for other benefits they offer, along the lines of NVH, high rpm capability, efficiency, etc. Now could they have chosen a nominal clearance and a tolerance for that clearance that would definitively result in significantly fewer low mileage failures? Perhaps, but there is nothing more than speculation to support that presently. Until there are some (more than a couple...) builds with fully blueprinted specs known for clearances and those engines are monitored for wear or lifetime, we'll continue to be mostly in the dark and speculating.