The phenomenon should be due to the reverse leakage current of D2. In fact the reverse current in a real diode isn't 0 as in an ideal component.
But there is something not clear to me: the 1N4148 diode should exhibit a reverse current of about 30 nA at 25 degC of temperature and 30V (it increases very much with temperature and voltage). This means that across a multimeter (we can estimate an input impedance of rouhgly 10 Mohm) we will have 30nA*10Mohm=0.3V and across a 1Mohm resistor (neglecting the 10Mohm parallel) a voltage of 0.03V.