About the expected accuracy about SFA of core, I suggest you to refer to the recent paper I presented to Sampling 2012 Conference (http://is.gd/ROeOYa).

The overall measurement error (including sampling error) is so large than the use of CRM cannot give you the accuracy. It will just give you the analytical error/bias which is a small part of the overall error.

Assuming you are doing the assay on 1/2 core, assaying 2 1/4 cores may provide a measure of variability, particularly when you compare that variability to the nugget effect of the downhole variogram for that hole. If this is a vein type deposit you may not have very many pairs in the downhole variogram. If the samples are channels the channel variogram may also provide some information on the nugget. It is also worthwhile to look at the chunks on the screen which go into the screen fire assay. If the chunks are actually visible and there are less than 30 or so then the marbles in a jar statistical approximation may also yield some insights. If you can actually see the gold in the core you may also be able to come up with the average weight of the gold chunks, calculate the number of chunks in the ½ core and again use the marbles in a jar approximation.

How much of the coarse fraction is being sampled for the assay? I wouldn't be surprised if bias/scatter is happening at that point. Best to use a rotary sample divider so long as there is enough material to be split among the eight receivers "equally". If not, then the nugget effect will still be evident.

The best way is to do a combination of screen metallic’s, size-by-size assays, and triplicate analyses of properly sampled coarse fractions. This is a more practical approach and can be augmented by a simple gravity test to reduce the cumulative errors of sampling, sub-sampling, losing gold in pulverizing and subsequent measurement bias.