You linked to a Power Meter City post. Here’s their detailed instructions on zeroing torque: How to Zero Your Power Meter - Power Meter City
Note that there, they recommend orienting the Assiomas at 12 o’clock (and the Vectors at 3 o’clock).
The Garmin Vectors also needed to be oriented – at least, the first generation, with the pods. I confess I’m not as familiar with the podless versions.
Pedal power meters need to know when the pedal forces are tangential to the pedal circle and when they’re radial (for example, pressing straight down when the crank is at 3 o’clock produces different vector forces than pressing straight down when the crank is at 6 o’clock). The Vectors were named “Vector” because they had to figure out the force vectors in different orientations. Pedal power meters now use accelerometers to determine their position but since they also use the accelerometers for cadence you can see that they’re a little “noisy” if you try to use accelerometer cadence to calculate gear ratio. (This is also an issue for crank spider power meters that have the option to use either a reed switch or magnetless accelerometer cadence – the reed switch cadence gives clearly more accurate data than the accelerometer cadence).
Crank arm power meters absolutely do need to be oriented since they (typically) don’t have a full rosette of strain gages (the exception is the Infocrank). And, as I mentioned above, even early crank spider power meters needed to be checked in different orientations. They typically don’t need that now since they use more strain gages now than they used to.
Getting back to Assiomas, they claim they don’t need manual zeroing, but I’ve seen data files from them (admittedly, on fixed gear track bikes) that are cleaner when one does a forced manual zero.