Lactate threshold detection is ridiculously incorrect

Did you know your real maximum HR? The real one discovered at a maximum effort test with a physician. 

Are you always running with a chest strap?

Yes, I know real max HR and always run with chest HRM.

First of all it is an estimation based on algorithms that may be very close to the truth for some but may fail for others. Secondly they are based on training activities where the goal is to improve prior to a race. A real test where levels are measured is a different story and is closer to a race thatn training.

What can be done to improve the results?
I'm not 100% sure about LT but VO2 is only measured and updated when using the standard Run activity (and Walk, but with a lower priority than Run), not with the Trail Run activity.
Reason is that it requires a steady pace on a flat course to measure V02 using the First Beat algorithms that Garmin uses. If you wnat the values to be as good as possible I would recommend that you use the Trail Run activity for everything except a standard run that you do frequently. The run should be flat (a easy run) and with a steady pace. If you know you can do 4:15 that's where I would start.
You can make a copy of the Trail Run activity if you want the same data screens as the running activity.
The down side is that you will loose other metrics such as Training Status.

I remember that my F3 always asked if I wanted to update LT values after a run, or keep the old ones. Can't remember I've seen that on my F6? I wish that feature for both LT and V02.

Finally; there are a lot of threads about GPS accuracy and distance being short. I don't have that problem but try to measure your "standard" run and compare with the watch. Pace is based on distance and time and we can agree that time is accurate...

What is your HRMax?

Were you getting the LTHR from the guided test, or during regular intervals?

Garmin's race prediction times are based on VO2Max and HRMax - LT isn't involved.

min LT is pretty close to reality, but my race predictions based on my VO2Max are much better than I can do.  

Thanks for your opinion! I never use "trail" mode, all my runs are on flat surface with minimum input from hills. Except this, I use stryd for distance/pace detection and it is extremely accurate, I see this when I run on indoors track.

Actually, this issue is not about accuracy, it's about Garmin providing garbage. More than 20% error makes data meaningless. 

Measuring LT pace of 4:46 for someone who runs more than hour with much higher pace is just physiological nonsense and very serious defect of LT detection algorithm.

There is a pace that you can maintain without problem for 10k or more, a little faster and your muscles can't keep after a few minutes. That's the point of the Lactate threshold. It's difficult to find.

You need a lab test to know your LT and say that your Lt provided by Garmin algorithm is crap.

My sleep tracking is crap, but my LT is spot on!

My HR max is 182, my LT pace is around 04:04, everything measured during tests and in a lab.

My LT and vo2max are wrong both, so I think that in Garmin's algorithms they are somehow related to each other.

I never use "trail" mode, all my runs are on flat surface with minimum input from hills. Except this, I use stryd for distance/pace detection and it is extremely accurate

This is important info since it will rule out things that can cause errors.

How does the following statistics look like for you?
https://connect.garmin.com/modern/report/49/running/last_six_months

Since I mostly do trail runs I don't have that many points in the report but they are all within reasonable limits. For me I think that the within reasonable limits but unfortunately it doesn't seem to be accurate for you. 

Just as a test; What happens if you skip the Stryd and use GPS for distance and speed for a couple of runs? Just to verify that there isn't a bug when a foot pod is used.

I know pretty well what LT is and have measured it a number of times in labs and during field tests.

Anyway, saying that LT pace is X when there is clear data that person runs at pace Y for more than an hour and Y is much faster that X, means that algorithm itself has serious problems.