Interesting.
I don't have a whole lot to offer from your observations - mine have been similar. What I'd say is that I've never really seen awful performance from "je", "bw" or "olap" operators - since these are associated with Join/OLAP Engine, all they really ever do are joins, aggregations, etc - but nothing too tricky. Of course, there may be room for optimization via filters, correct modeling of join types etc - but most of the time, performance hang-ups have been with CalcEngine operations. So I've just become accustomed to looking there.
However - I'm now rather uncertain about the latest optimizations in HANA. Once upon a time - it was a fact that calculated fields that had formulas which referenced dimensions were calculated in the CalcEngine, after aggregation (in OLAP Engine). In fact, the studio would give you a warning that "Calculate before Aggregation" would be ignored in cases like these - meaning the OLAP Engine wouldn't handle such calculations.
However, I now just create a very simple Analytic View with one dimension table, and I created a calculated column with an if statement, which referenced a field from the dimension table - and I checked "Calculate before Aggregation" - and the VizPlan shows everything being executed in OLAP Engine (i.e. operators with "bw" or "olap" prefixes). See below:
So I'm not quite sure what to make of this. In general, I'd say the usual recommendations apply - avoid calculated columns when possible, create them as generated columns where possible, etc etc - but it looks like I need to get up to scratch on HANA's latest nuances. 
Weird about Explorer as well. I haven't done much with Explorer before although I have seen problematic SQL generated from Lumira. But that's another story. Keep me posted on what you find!