the problem is that
* we don´t know whether you are talking about a real circuit or a simulation
* you don´t tell about the load
* you don´t tell how much voltage you see at the output
* you just talk about 5V control signal when ON, but you don´t tell what´s the control signal when OFF
* we don´t know which resistors you used at the control signal
* and how you get your voltages
***
Thus I can just guess
* a defective part
* a measurement error
* wrong test conditions
* wrong expectations.
+++
My recommendation:
* give complete informations
* focus on the "faulty situation"
* measure V_GS of each MOSFET
* measure V_DS of each MOSFET
If you are measuring with a DMM, no output load, you will read a voltage
because of this -
There is still some drain current. And this does not account for leakage in the NMOS MOSFET Idoff
current at its Vgs = 0. Which can cause a non zero Vgs in the PMOSFET.
Aside from all the other issues pointed out here, 100K seems awfully large to me for the (presumable) gate pull-up. There might be enough leakage in your circuit to keep that MOSFET turned on.
10uA*100K means -1V Vgs, not zero Vgs, on the PMOS. You
can substitute whgatever leakage current you measure (or
simulate as the case may be) on the NMOS drain.
As a rule-of-thumb, put a 1k or up to 10k resistor ground at point A. As danadakk mentioned, it might well be the 'off-state' leakage current which causes a 'phantom' voltage, very typical. Datasheets will state leakage current and can presumably calculate Rleak better than just the 1k to 10k habitual guesstimate.