Let me try to explain this part. The chips you are refering to are AMC7135 current regulators. These accept voltages from around 2.7v to 6v and what they do is only allow around 0.35A (350mA) to pass through them. The mini has 3 channels controlled by the MCU. Channel 1 uses a single 7135 chip as seen in the pic below.
This is pretty low current so it doesn’t produce much heat and the brightness range is small. It makes for a nice moonlight level and up to maybe 150 lumen.
Now channel 2 uses the 6 chips plus the 1st one. So now you have 7 chips that allow up to (.35 x 7) 2.45A. This is what the top of the ramp is set to. So all 7 chips do produce some heat, but you can see the ground tab is soldered to the outer copper ring and that is pressed against the battery tube. That is your heat path (red arrows). This looks pretty decent. Actually, the battery tube does not contact the driver so that heat is not being drawn away.
The temperature sensor is built into the MCU on the other side of the driver board as seen below.
The way heat gets to the MCU is through the 8 metal legs.
There is definitely no thermal paste on the 7135 chips as they are soldered.
On Turbo, all 7 chips are not used and all the power goes through the FET. You can see there are a lot of vias around it to spread the heat to the other side of the driver.
I really don’t know if this quick step down from the top of the ramp is a standard problem of the 7135 chips getting hot and heating the MCU much quicker than the body or if something strange is going on in the software.
I believe the 7135 chips may not have to shed as much heat at lower voltages. If so, then that might explain why the lower voltage allows for longer runtime before stepping down. Maybe there is some unknown thermal path between the 7135 chips and the MCU that is causing the problem and this mystery path is not between the FET and the MCU which is why Turbo does not step down early.
IDK, it’s a bit of a mystery.