The lens narrows the the beam down to a point but the other side of that point it just keeps widening indefinitely. The beam is not collinear after the lens. See the schematic side view in the intro topic to visualize it. Even if it reflected off the bed perfectly, it would be spread out enough by the time it bounced back not to burn. You can try engraving cardboard rather than paper to see that it’s coming from above and not below, if you like.
It’s quite common for the beam not to be centered in the tube and to bounce off the edges of the nozzle.
See the videos that American Photonics made to describe their 3rd mirror alignment tool which demonstrate why it’s easy to have the beam not concentric with the tube.
But also, have you checked for belt tension and for play at the wheels?