Acoustic, there is always a lot of discussion on the topic of anchor performance - and rarely agreement. The basic problem we all have is that anchor performance out in the real world can not be empirically modeled, and so those tests - while perhaps outlining the issues thoroughly - probably don't offer hard/fast guidelines for what's "best" (as tho' there is such a thing).
Your aluminum spade is an anchor designed & built to be light (for a given level of performance), yet what you want is ultimate holding power; it's no wonder you are a bit dissatisfied. You'll find Alan (the Spade 'inventor') and Peter (ditto for the Rocna) are both very persuasive in selling their 'next generation' anchors (I just love how that phrase 'positions' us to think about old vs. new and good vs. bad) and they are both seasoned cruisers...and yet, in this thread as elsewhere, we find that people are experiencing the same basic issues with the performance of those 'new' anchors as the old standards. Out in the anchorages (see TrueBlue's comment above) we can see different boats having different experiences with their ground tackle and it's easy to attribute it to different anchors...but in truth, the variables involved are almost infinite: varying seabed, boat windage, different bottom contours, wind & wave patterns within the same general area, varying scopes and with varying lengths of chain and, most important of all, varying skills in choosing the anchor spot & setting the anchor in the first place. While you can find test results that might compare the anchors you are asking about, it's hard to know what that will really mean - for you, or any of the rest of us.
You seem pleased with your Bruce's performance, yet are willing to spend meaningful sums to have 'better' performance. Why not stick with what your experience tells you works for you and, if you must buy more 'insurance', move up to the next size Bruce? One thing that has been generally validated in all the anchor tests I've seen is that the performance of burying type anchors (CQR, Bruce et al.) is more than linearly improved by using a bigger model.