hosci,
yep - full = only a bit of derate / empty = a lot of derate
Will make no difference in the end - as j_t and OzExpat have more than hinted at, you don't readily know how much the acceleration segment on one engine is gonna take out of your total climb gradient. As the mass varies, so will the limiting obstacles - on a computer in an office, or with a GWC on stand, it's all very easy to work out.
Re. question 1, from start of 2nd segment.
Re. qestion 2, if you insist on using the SID procedure design gradient, you need to check the "Horizontal distance to end of 3rd segment" graphs in the AFM first. Your limiting mass will be the one that places you at the end of 3rd segment above the PDG. Secondly, you will have to check the final segment climb + the en-route climb limits to be at or above the PDG all the way up to the MSA or the end of the SID.
However, due to increasing obstacle clearance (obstacle identification surface + 0.8%) & a much wider survey area (thereby possibly including even more limiting obstacles) - you will end up with a much lower obstacle limited TOM than a GWC could provide you, So why not forget the SID & go with the GWC + any emergency turn procedure stated therein?
Brgds,
Empty