As Hippy says, PULSIN is not a good tool for detecting missing pulses. It's intended to measure the length of a pulse, and is fairly patient about waiting for that pulse to arrive, so it can never be quick telling you the pulse never does arrive.
As you are producing a stream of thousands of pulses per second with PWM, you might use COUNT instead. You choose a window of time, which can be very short, and it counts pulses arriving (if any). You could set up a window of say 5 expected pulses duration. If the count result is in the right area, say 4 - 6, the beam is "good". Outside of that, it's broken.
The advantage is that detection time is as long as the window you choose - which can be considerably shorter than PULSIN timeout.
You can tweak response to an extent by altering the window period, and the leeway you allow in expected pulse count, to allow some immunity to dust motes and flies, while still detecting cats or superheros etc.