Background masking philosophy

[Mason]

I've been doing a lot of manual background masking of my photo sets before handing them off to PhotoScan.  I find it makes a big difference in the results but is rather tedious.  So my mind wanders and I think of questions:

Masking borders
I'm rather sloppy around the borders of the object that I'm masking.  I'm careful to include all of the object but I'm not too concerned if a bit of the background appears as well.  However, I'm just assuming that it's not worth the effort to be more precise.  Am I wrong?  Should I spend more time to be careful not to include any of the background?

Object occlusion
In most cases, I've been walking around an object set on a table, taking pictures with a single camera (using a tripod when I can).  For low angle shots, sometimes the table itself blocks part of the object.  My assumption has been to consider the table part of the background and mask it out even though in this case it's partly in front of the object.  Is this correct?  The only reason I can think of to include blocking items with my object (and guess the space of my object behind it) is if PhotoScan is using masks as object volume hints.  (Which would seem quite helpful to the process.)

Fixed camera arrays
For those lucky enough to have lots of cameras pointed every-which-way and can take advantage of background differencing to generate masks:  How do you account for yourself in the photos?  Do you mask those images manually?  Don't bother masking them?  Do you hide from the cameras?!?

Thanks for any thoughts....