The analog Tq meter and the atpcs system get their Tq information from Tq sensors in the engines. If the sensor gives incorrect information this may at least in theory activate the ATPCS and thus the autofeathering of a fully functioning engine. In the FDR picture that I have looked at (
http://pbs.twimg.com/media/B9JarHTIAAAgAy3.jpg:large) it seems that the torque of the autofeathered engine was norman until the feathering started. So it seems that the torque has not dropped and thus activated the ATPCS cycle. But there are two differend methods to measure the Tq (I don't remember the other one precicely but it is calculted by the engine control system) and the one that is visible in the FDR tape may not be the same Tq information that is used by the ATPCS.
In ATR MEL it is a normal procedure to turn off the ATPCS system for takeoff and use 100% Tq if the analog Tq meter is not giving correct information. This is because there is a risk that the malfunction is due to broken Tq sensor and it might activate the ATPCS in a fully functioning engine.
If you look at the ITT of the engine #2 it stays at about 400C after the feathering. The engine #1 drops quite quickly to around 150C after it is shut down. To me it seems that the engine #2 was just running on idle and feathered all the time. But power lever of engine #2 was at full power all the time. I have never tried what happenes when you feather an engine with full power in ATR but there might be a logic in the EEC that turns the engine power to idle or near to idle despite that the power lever is in full power.
The situation was most likely confusing. At least in our company we have never simulated inadvertent autofeather in simulator. I don't know what happens in such situtaion.