uCurrent Gold: offset + noise vs. original uCurrent

[MysticPixel]

I was excited to receive my uCurrent Gold the other day, so I swapped it into my rig (kudos on using screw terminals w/wire holes for all inputs! I got sick of making little adapter doodads) ... but I'm finding some very strange results.

First off, note that this is a direct swap - nothing else in the setup has changed, I took these two captures within minutes of each other and made sure not to disturb anything else (nearby wiring, potential noise sources, etc.) The leads into the uCurrent vs. uCurrent Gold are identical, I just unscrewed them from one and dropped the other in. I am sampling with a TI ADS1015 ADC eval board, with the inputs configured as differential to allow for high-side measurement.

Range is 1mV/mA. ADC settings are identical. See attachment for the differences (note that the vertical scaling on the second graph is 2x that on the first, so the differences are even larger than they appear)
In short:
* original uCurrent has minimal DC offset, noise under 1mV
* gold uCurrent has -4mV offset and is really noisy, it jumps back to 0mV. These peaks are repeatable across tests and occur just under 40 per second.

I'm not ruling anything out at this point but I wanted to post here and see if anyone else had some experience with comparisons between the two.

[theatrus]

The Gold does have significantly higher bandwidth, which can account for the higher "noise" (depending on your load)

[EEVblog]

The Gold does have significantly higher bandwidth, which can account for the higher "noise" (depending on your load)

That is likely it. We are talking 100x the bandwidth here.
The output offset though, I don't know what's going on there. I have never seen one that was out of spec. 
This is a typical offset/noise distribution graph