When the reading is wrong

Stuck, or wandering

Pick what you see. The red LED helps only at the top: it lights once SIGNAL passes about 2 V, so lit proves the sensor, VCC and GND are all working. Dark proves nothing in an ordinary room, where the reading moves well below the LED's threshold.

Stuck at 0

No current is reaching the 10 kΩ. The usual causes, most likely first: VCC not wired, which looks exactly like a dark room; SIGNAL on another pin than the sketch reads, or on NC beside it; and a light with no infrared in it. A phone torch is a white LED, and to this sensor it is nearly dark. Test with a TV remote held to the dome, a button pressed: the reading should jump.

Stuck at the top

The sensor has run out of room: SIGNAL is about 0.4 V under VCC and more light changes nothing. Sunlight does it, and so does a filament or halogen bulb, or an emitter held against the dome. The red LED is lit. Shade the sensor with a short black tube, turn it away from the window, or move the emitter back until the LED goes out.

Wandering

This is the block working. It adds up every infrared source it can see, and nothing on the board filters any of it out: a remote pressed across the room, a cloud past the window, a warming bulb, even the sensor's own temperature, which the datasheet shows raising its current by about half a per cent per degree. Average several readings for a steadier number. For a beam, read the room with the emitter off and subtract, as a beam that breaks does.