Turns out I was using the app GPS Test by Chartcross Limited (space) but you are using GPSTest by barbeauDev (nospace). I got GPSTest (nospace) going and my Pixel 4 is using L1 sand L5.
But as you say it does not seem to actually improve anything. GPS Test stops getting lower at 3m and never better. GPSTest goes down to 3.8 and never any better. This could just be Android reporting a min of 3m ever. I read somewhere once it will never actually report better than 3m in the software (liability?).
Both Collector and 123 are both reporting 12.4 feet - so right at the 3.8 m.
Hope that helps?
See this article for more info
https://medium.com/@sjbarbeau/dual-frequency-gnss-on-android-devices-152b8826e1c
Other quote
"There is a bit/lot of confusion because chips report “theoretical” accuracy and that’s now under 1 meter with DF, however no one’s seen anything better than 3/5 meters I believe with actual testing. It’s not hard to see why, tiny GPS antennas, interference, etc…DF on smartphones is going to be mostly useful on smartphones to eliminate “bounced” signals in deep urban canyons."
In my last job we did a lot of testing and in the end it had way more to do with the size of antenna then anything else. This is why a bad elf is no better for example. Once we added a antenna the size of a dinner plate it got way better.