I'm afraid I wasn't being sufficiently clear in my questions and concerns regarding the offsets and how to deal with them. I will try to add clarity with this post.
The Directiva 1 design is a two-way design in an off-the-shelf box that does not time align the woofer and tweeter. The tweeter is above and effectively in front of the woofer. From directly below the speaker, you would expect the woofer energy to reach you before the tweeter energy, because of the vertical offset between the drivers. From in front (on-axis), you would expect the tweeter energy to arrive slightly before the woofer, because the woofer is effectively a little bit farther away. From directly off to the side (left or right), you would expect the energy from the woofer and tweeter to reach you practically simultaneously.
In order to correctly simulate the response at any location in space, the relative distances to the drivers must be accurately represented. (The impact of errors here will obviously depend on the magnitude of the error.) In order for this to happen, there must be synergy between the delays introduced in the simulator, and any delays captured in the data files themselves. If the data files captured a delay, then the simulator needs to avoid introducing that same delay, or else it will double that delay and be in error. If the data files did not capture a delay, then the simulator needs to account for that delay, or else it will be in error.
I went through the trouble of turning spectral plots into time domain (impulse response) plots. For clarity, I decided to use distance (mm) on the x-axis. The measurements were apparently taken from 3 meters away.
Let's start by seeing what happens on the sides (left and right). I don't know if there's a standard for which is "left" and which is "right". Would it be from the speaker's perspective or the listener... at any rate, it doesn't matter for this analysis. Left or right should show a roughly equal distance to the tweeter and the woofer.
The tweeter looks a lot more impulsive at this angle than the woofer, but we can see that they start to respond to the energy at around the same time (and distance).
Now let's look above and below.
There's a large vertical offset between the woofer and tweeter, so we might expect to see a large difference between the alignment here, but we don't. The woofer and tweeter still both begin to respond at about the same time / distance. This implies that the simulator will need to account for the vertical offset.
Let's see what happens on-axis:
Both the woofer and tweeter are much more impulsive at this angle (have much more high frequency energy than at the other angles). Interestingly, there appears to be a delay associated with the woofer, when compared to the tweeter. The files appear to capture the fact that the woofer is farther from the mic position than the tweeter is. This implies that the simulator needs to not account for the z offset between the drivers, if it wants to represent the natural physical layout of the drivers on the baffle. However, if the system designer wants to manipulate (or compensate for) the natural physical layout, they can do so.
I believe this is why Rick had a delay in his active crossover, but had a z offset of 0 (none) for his tweeter and woofer. The z offset was apparently already accounted for in the files (even though the Y offsets weren't... which is a strange inconsistency and makes me concerned that everything may not be fully understood by all involved, myself included). To be clear, the crossover delay and the z offset are not the same thing at all. The crossover delay is a fixed amount of time regardless of mic position. The Z offset generates a delay that is dependent on the microphone position. By aligning the response on-axis, Rick has effectively misaligned the response at other positions. This is a design decision and the trade-off was apparently worth it.
For anyone simulating passive crossovers, it looks to me like the best thing to do will be to make sure you have the appropriate vertical offset for your woofer and tweeter, but don't include a z offset, because that's apparently captured in the files.
Someone please correct me if I'm mistaken.