Well, the LO energy has to switch the diodes on and off. It has to switch both diodes equally on or off to act in a balanced way. If you have the two diodes in series between ports 2 and 4, it is not clear to me how the LO energy can turn the diodes on or off. Current has to flow thru both diodes to turn on, but the voltages at port 2 and port 4 are only 90 degrees out of phase with each other--not an efficient way to apply LO energy to the diodes.
A better way would be to place an RF shunt capacitor in between the two series diodes. One lead of the capacitor is connected to the series connection of the diodes, and the other lead of the capacitor is grounded. That way, LO energy appearing at port 2 causes current to flow in the diode connected to port 2, independent of the other diode.
Similarly, LO energy appearing at port 4 causes current to flow in the diode connected to port 4, independent of the other diode. In this case, the capacitor has to be large enough to be an RF and LO AC ground, but be high impedance at the IF frequency. You can take the IF signal off of the capacitor.
A different circuit would actually connect each of the diodes independently to ground, breaking the "series connection". The IF is taken off of either the RF or LO port thru a lowpass filter. The LO and RF port use a series capacitor that is large in impedance so that you do not lose IF signal out of either of those ports.
I am not a mixer expert, but what I am trying to say is that unless you get enough LO energy into each diode to switch on and off, neither circuit will work. You non-linear simulation program is smart enough to know that, and will give you poor results.