I'm assuming by an ATP plant you mean something like the one at http://is.gd/eywnDa? So I'm assuming the intention is not much crushing (perhaps a bit of sizing?) then washes with water to recover the gold.

One of my jobs as a student geologist doing some industry experience for 3 months was dealing with this sampling problem for tin exploration. We were collecting 40 to 80 kg of material for 12 inch cable-tool bore holes and washing them through a sluice box to concentrate the tin. Relatively easy to set up with a hopper and two different screen sizes at one end and hose at one end and a three meter length of steel sluice box (say 20 to 30 cm wide) with a two riffles, one at either end of the sluice. With person-power and couple screen sizes at the front end have two people washing the rocks and breaking up the clods with a hose. In the sluice box have a person with a flat blade agitating the fines against the current between two riffles (one at either end of the box). The first riffle should be placed after the point the steam flows into the sluice box. This riffle will capture any lumps that get through and causes the water flow to be laminar between the first and the second riffle that is place about 50 cm in from the other end of the sluice. The second riffle is designed to capture the heavy’s fines. Both riffles are just square profile woodsticks (say 2 cm square) that can be jammed into position and tapped out easily for cleaning the sluice.

Once all the fines are washed into the sluice between the riffles, keep washing until the fines agitation only produces is clear water. When you get to here, stop the water and pull out the last riffle collect the heavies from the sluice for panning. The concentrate can then be processed at the lab to estimate the recovered grade but you should be able to get a photo if you have visible coarse gold and make a visible estimate especially once you start getting some lab data back for the concentrate.

My thoughts are a knelson concentrator will get more gold than a sluice but the hand processing is likely to get more gold than a trommel . You may also wish to dry and crush and split the coarse rejects to get an idea of how much gold is in this fraction. If you are concerned about the sluice reject you could take some time increment cross-stream samples from the exit stream into a large calico bag; let these drain and send them off for assay.

In a similar situation, we performed close order hand auger sampling with 'bit and bale' type augers. With extensions and lots of manpower, we were able to reliably collect samples from as much as 10 meters. The individual 1' samples were immediately bagged in entirety.

We had access to a pug mill and used it to break up the lumps. With a sluicing arrangement much as described above, we were able to effectively recover a very wide range of gold particles. We checked the 'tails' of the sluice regularly with a handheld White's metal detector followed by gold panning for verification of gold recovery.

The differences charted in recovery in the sluice vs. hand panning was charted and correlated with the character of the regolith. In our case, the 'regolith' was a well-developed laterite, and below was a saprolite from which the majority of the gold values were recovered. We found virtually no gold in the laterite except where relict veins/dikes were present as distinct, saprolitic bodies in the laterite.

This was all pretty much in the second-pass phase of the exploration program. Based upon the results of this work, the program transitioned immediately into a routine collection of samples which were then transported to a lab in Australia with the facilities to handle the entire sample. Good luck. "Regolithic" environments can be fun to work with.