Just a shot in the dark here, but I could see simple signal processing and delays being an issue with power match. If you take the Drivo II trainer, for example, it supposedly takes a power reading hundreds of times per second. The resistance it needs to apply is constantly changing, since wattage is a function of your pedaling cadence and smoothness. Our legs aren’t perfect, so it’s not like the trainer just “sets” the resistance and stays there, it fluctuates that resistance constantly to compensate.
That said, without an external power meter, you still get up and down power fluctuations… sometimes only 1-2 watts +/- if you’re really smooth, sometimes it’s 10 or 20w if you’re in the large chainring or are fatigued and not pedaling super smooth. From what I have read on DC Rainmaker, Ant+ FE-C (not sure about BLE) Erg mode works by the controlling device (in this case TrainerRoad) sending the trainer a given wattage to hold. The trainer then holds that wattage until it is given another command; TrainerRoad does not constantly send updates telling the trainer what to do.
Now, you add an external power meter. That power meter is communicating directly with TrainerRoad via BLE or Ant+, processing that signal, comparing it to what the trainer is reporting, and then has to constantly send the trainer updates for wattage. That adds a delay, both for the sending of the data from the power meter to the the phone/computer, TR to process it, and re-send a new signal to the trainer, the trainer to interpret that command, and apply the resistance change.
On top of all that, you can have signal interference which causes delays, micro-drops, and packet loss. You can see this by uploading a Zwift log (if you Zwift) to Zwiftalizer and scroll down to the signal section. It shows you a percentage of data loss, signal strength, and how many times a connection was lost to any of your devices.
From Elite’s customer service: “…you have to consider that the software itself has its “time” (it must acknowledge that the first interval is finished, verify the power of the following interval, calculate the resistance command to send to the trainer and finally send the message). Also, the communication itself has its own delays (due to interference for example).” So they claim that for the Direto and Drivo, intervals under 10 seconds may be too short, due to the all of the above (and actual physical adjustment within the trainer to control the resistance). Adding an external power meter to constantly correct the trainer adds a whole new level of complexity and delays.
I could be way off base here, but I bet this definitely plays a role for some users depending on their specific conditions of devices, trainers, and connection stability.