At last it works!
I've redraw the schematic above. Now it's even more simplistic.....
I was obsessed in getting 5v for the STROBE signal but I've found that it's more important to get the 0v. Precisely because is that state what initiates the reading process.
The Neos still sends movement with only 2,2v and 0v.... But for compatibility with unknown devices(MSX mouses, Arduino adapters, etc) I've managed to reach 4,70v.
RTS unloaded oscillates between 0v and 12v, but with the original Boxsoft(OR gate) it drops to 7,7v. With the present project(transistor) it even drops more to 5,5v... Could it be harmful eventually for the EP? I think not, it is protected inside with a 10K resistor.
I have to do one more test with this "Optoboxsoft" adapter. In the past I have experienced errors of direction with some games when playing with a joystick plugged in Control2 when still connected the original Boxsoft adapter and Neos in Control 1.
I haven't my joystick at hand... a pity. But it must work(theoretically....).
Also, now is easy to make work an auto-fire joystick on this new adapter. Except for the Fire button moved to pin5 of the D9 connector, it is a real Atari type connector with 0v as common.
No more excuses for not having mouse on the Enterprise with this easy and cheap project.... if you find a Neos mouse on Ebay.
Now I can continue with the Arduino adapter...