rising and falling edge detection problem..

More information is required to answer your question fully as you don't say whether you are trying to trigger the monostable on both edges or just one edge.

If I understand correctly you are trying to produce a shorter pulse from the 20ms square wave (as suggested by your list of timing component results)? If so, then I would suggest that the 74HC123, or any family "123" device is not the best to use for the task. This is because it is a re-triggerable monostable and thus any noise spikes present on the 20ms signal could re-trigger the timing generator before it has reached the full timing period. This will result in a longer pulse than expected for the theoretical R-C values chosen.

A better device may be the non-retriggerable type 74HC121. This will give a fixed pulse length for each trigger event, however, do note that a noisy input signal can still trigger this type repeatedly as well. If the signal is very bad a large R-C filter followed by a Schmitt-input gate may be needed to clean up the 20ms signal, before it is applied to the monostable input.