Spectre modeling of inner finger devices within big abutted row

hello experts,

this could be a mixture of process vendor or EDA vendor question. I'm betting my luck here. 

I have some critical devices that has to be accurate and matching. Referring to my diagram, I have shown an example of two of them - transistors A (4 fingers) and B (6 fingers) where each finger has the same W/L. When I specify the transistors using only fingers (=4 or =6, with m=1) I can see that each finger doesn't match so well and I believe what's going on is that the models know that the outer fingers have "issues" due to end effects. If I layed out the transistors separately as shown on top I will indeed get a mismatch between them just as the models predict. However, I have many transistors on this bias line and my intention is to lay them out as shown on the bottom and in that case they will match very well because those end effects are far away. 

In order to get my simulation working well for matching purposes, I've had to resort to e.g. using a multiplier of 4 instead of 4 fingers. that can get me perfectly matching "fingers" but then its modeling additional effects which doesn't really exist in real circuit as laid out. Is there a way to use fingers and somehow instruct the simulator or the models to ignore those end effects?

thanks,

David