We would be discarding a lot of weights, but we definitely couldn't use most of those for BeyWiki anyway, so as much as I would like to include them for archival purposes, it would be a waste of time, yes.
There are some cases where the weight alone can tell you the mold (some SonoKong versions, some very distinct ones like Vulcan's tends to be, Mold 1 Earths, etc) so those could be included with a note on the 'safely assumed mold'.
Of course until we have a clear guide to mold differences, some are *very* hard to tell apart - for example Burn Mold 2/3 is close enough that I'm not yet convinced it isn't just a result of denting/wear, as I shoud have a mix of Molds 2 and 3 and they all look very, very similar, with noticeable 'denting' in the lines being the main cause of that, so I cannot tell what molds mine are. (I've posted Pics over in Beyblade Random Thoughts where similar discussions are taking place).
The thing I worry about is with older weights, people might not know that they have say a mold 3 burn rather than a mold 2 - now to be fair, if I have the expected mix of Mold 2's and 3's, then the weights are very similar, but odds are there are cases with other wheels (this is why I advocate including the full source - brand and release - for any reasonably weighty part that a weigher can).
There are also likely to be cases where an abnormally light or heavy weight for a mold gives the idea that mold is much lighter - for example, 0 cylinder Rock. Beywiki lists a very light weight for it which was at the time a safe assumption because no one noticed the metal was moved to the underside rim until I looked into why my 0 Cylinder Rock was very close to being my heaviest recently - I suspect very strongly Takara Tomy aimed to keep the weight consistent. However, an average would indicate it being lower - while it's unscientific to argue with data, it's not completely out of the question with very small datasets.
Then again, that is something one could just explain in the article, so I am probably finding problems where there are none.
Tables thing:
We have to expand that to include weights of different molds anyway IMO, so splitting off weight and dimension tables might be good anyway. Weights table could include Average, Min and Max, with a separate table for each mold, which aside from the original mold could be put in the Mold Variations section. For parts with subcomponents, we continue to add the two subcomponent averages as we generally do from what I gather (this is less accurate than having everyone re-weigh such parts whole etc etc but it looks better and is less hassle), and I guess use the total of the mins and maxes of the subcomponents for those figures.
Would say to include n (total number of samples used to generate that average) however for parts with subcomponents, the fact people rarely do more than a total weight means resulting disparities would look very odd, unless we discard an absolutely absurd amount of perfectly usable data.
Take for example Phantom - we've got a tonne of weights for it thanks to this thread, but only three I've seen have separate PC Core and Metal Frame weights - Mine and both of Byser's, all of which are a little over 42g total (i.e. bottom of the weight range). Given the core weight tends to be within a few tenths of a gram (and I doubt there's going to be much variation there), I figure the most accurate result will come from subtracting the average core weight from the full-wheel weights, and using those numbers along side the actual Metal Frame weights that we have to generate an average - though as I think I said in the errors on the wiki, I'm going to wait for someone to give me the go-ahead before I do so, as I'm not 100% confident in that. But anyway, that leaves us with N=3 and N=10 for the core/frame respectively which is kinda weird.
But yeah, Average, Min, Max, and if people want to see more they could check the compiled weight document when it's done.