sorry for huge delay.
Although was bad, the precedent circuit did work because when input was 3.3V the 47k resistor prevented the pnp to get into saturation zone so as the rgb stip common isn't gnd but 12 V, current prefers to flow to ground through the load and pnp is not compromised.
The follows circuit probably would be much correct but for sure much slower shutting down led strip.
Another solution proposed by a friend of mine is to use optoisolators and OP-AMP, but since i want to use discrete components I have rejected this solution. Please post your considerations.