It’s important to note that power is exponentially related to speed, so you can’t simply use average speed with that formula to provide average power (as any lower speed segments would have undue weight on the result). If, for instance, you rode half of a ride at 10mph (72w) and half at 20mph (262w), the average power would be 167w. If you simply averaged those speeds (ie 15mph) and calculated power from that, it would only give you 145w. Naturally, that’s a significantly exaggerated example and you likely had much less variance during a FTP test, but it is something to keep in mind. For comparison purposes, you’d be better off finding a few representative points in the ride and comparing the power value calculated using instantaneous speed.
With that said, the magnitude of the difference you noted is higher than I’d expect for something like this so there is likely other things at play. Just figured that I’d mention it as it’s likely a contributing factor to the issue at hand.