I'm pretty sure its because of the amount of stale and discarded shares, which are due to hardware errors. You also have a high amount of get fails at 10%. These figures are massively higher than I get.

I'm at a bit of a loss here to be honest, maybe someone else can chip in. The only thing I can see that looks like it could be causing it is that your temp is a little on the high side, these really need to be kept cool, where have you got the miner?

We keep ours in an unheated cargo loading area and they never get over around 42 degrees.

Thats alot of stale shares!

I have 4 ants mining now, 14 hours 9000-14000 shares each and not even 1 stale.

Are these numbers what gives an idea of the number of errors?

(17 hours 35 mn)
LocalWork: 2,870,281
DiffA: 2,050,944
DiffR: 29,824
DiffS: 729,728

upping this one last time in case some had an idea?

no matter stock or o/c'ed to 1400 it's hashing with a ratio (DiffS / Local Work) of 12-15% no matter either pool (only tried BTCguild and BTCdig).
API log shows very consistent numbers throughout the 16 miners (all within 1% of each other).

Local hashing rate (Cgminer status page) is on spec (200 stock) - (215/220 @ 1400).

How did you fix it by proxy? can you explain more?