Advice on the final volume levels for YouTube(with videos*Zoom H5*)

[Vindellama]

Ok...
So I tried to do a proper recording for the first time in several months.
I should do this more often to practice controling my stress levels.
This is the first time I've tried to use a proper video editor instead of windows movie maker.
I tested several...
All of them were increasing the gain a lot upon exporting. Making a perfectly leveled audio track change into a loud distorted thing.
I thought it was some sort of bug.
But when uploading one without a normalized track to compensate, the level went back down on YouTube. And uploading one of the distorted videos got it back to the original lvl without any distortion.
Could someone explain these changes in audio levels from exporting (flac/pcm to aac), and uploading?
I prefered one called Filmora.
Being an ultra begginer video editor and shooting on a medium range smartphone, it seems to do the job for now. The image quality on this phone is ultra contrasty and saturated in my usual playing enviroment, and there is one preset on Filmora that seems to reduce the artifacts a lot which I liked.
Also... I'm trying to push the gain as high as possible without distortion, since most of the music channels seem to have a higher final volume than what I've got in the past. And that makes easier to hear through a smartphone without headphones and such.

Here is the same video in crescent audio levels (remember to listen in 1080p)... EQ is flat, with some reverb added in post on Reaper to match the original song. I'm pretty impressed by what the zoom H5 can do with the stock mic capsule in the right position. There was an even better (soudwise) take on a slightly different position, but the playing was a bit sloppier and without the special guest. From previous experimentation. I liked it better when the zoom is positioned around the soundhole, instead of the usual advice of positioning it in the neck joint and turning it around for "eq".

Unormalized track pcm 32 bit(94khz?) to aac 320 kbps 48khz(original tracks in pcm 32bit unless specified):


Slightly below "normalized" level (maximum peak 1db?):


Normalized (maximum peak was around 3db?):


Pushing the gain without noticing distortion (maximum peak around 4-5db, source was flac):


Which one do you think have the better level for multiple plataforms?
I think maybe the usual normalized is the better one... When pushed to 4db it is slightly louder, but seems slightly less smooth?
Also...
From which format is better to convert into aac? flac, pcm, wav, etc?
Edit*
Now reviewing... Maybe the low-mids got too boomy with the reverb?