And there comes another confirmation:
https://connect.garmin.com/modern/activity/1087022623As you see the F3 doesn't miss a single length, nor did it with the previous software versions.
GFD importer shows 3.93 and
inconsistency in lap no. 1 after 0,00 km: missing length
inconsistency in lap no. 1 after 0,00 km: missing length
inconsistency in lap no. 1 after 0,00 km: missing length
I'm afraid the new code of the .91b messes up with something inside GFD. OMB, if you think the problem is in fix #39 perhaps a beta version without that fix might do the trick.
Meanwhile it's better to still hold our horses until this version is released officially because we never know. What I think is that because of 91% of those who complain being incompetent, 8% having faulty devices and 1% having magnetic problems, I'm getting screwed because of this glitch to save just that 1%.

I'm keeping all these .fit files in a safe place. Can't import them with the bug in the standard ST3 importer (time bug) and can't import them with GFD's missing lengths either (wrong data averaging).