As a researcher on starch and an author of the book Ancient Starch 
Research, let me add my own perspective on the issue of preservation, 
which is something that I am very interested in understanding.  I 
certainly don't take it for granted that every starch grain seen on the 
surface of stone tool or in sediment is necessarily 'ancient'; a case must 
be made by the analyst as to their reasons for interpreting the material 
as an archaeological residue.  To that end I have published a few papers 
that specifically deal with the argument of starch preservation and more 
broadly on our understanding of what an organic residue actually is - if 
not contamination, as is often claimed. Most Recently I published a paper 
looking at residues on museum artefacts and discuss the preservation 
taphonomy of starch in that particular context.

Barton, H. 2007. Starch residues on museum artefacts: implications for 
determining tool use.  JOurnal of Archaeological Science 34: 1752-1762.
 
In that study, I found that starch granules were often found in very 
different states of preservation on a tool surface.  Some granules were 
heavily degraded by microbial activity - showing all the charactersitc 
forms of enzyme attack that the literature predicts and fits with 
expectations, however, many granules, from the same residue extraction 
were recovered in a perfectly good state of preservation.  Apparently 
unaffected by enzyme decay. why?  I suggest that the formation of 
an 'organic residue' - which must generally consist of some mixture of 
mashed up plant material, lipids, etc. dries rapidly and entombs some 
starches preventing access to microorganisms - some starches are not so 
protected and are degraded by all the mechanisms known to degrade starch. 
Sometimes these old starches have different visual properties to modern 
starch, and have been altered in some way; most likely changes to the semi-
crystalline layers judging by the changes seen in the extinction cross 
under cross polarised light (see Fig 7 and Fig 8). Decay does occur, and 
starch are clearly removed from the environment - probably in very large 
numbers, but organic residues themselves help preserve starch - which is 
itself an extremely stable compound.  Everything we could find at the time 
that discussed likely mechanisms of preservation and degradation of starch 
was published in the Taphonomy chapter in Ancient Starch Research (Barton 
and Matthews 2006).  Old starch also stains with IKI and Trypan Blue as 
per our expectations for modern starch - which is also telling us 
something about the chemical and structural properties of preserved starch.

Barton, H., Torrence, R., Fullagar, R. 1998. Clues to Stone Tool Function 
Re-examined: Comparing Starch Grain Frequencies on Used and Unused 
Obsidian Artefacts. Journal of Archaeological Science 25: 1231-1238 

In this early study (largely in response to the criticisms of residue 
analysis justifiably raised by Roger Grace) I undertook a study with 
Richard Fullagar that involved a blind test comparing starch granule 
counts with assessments of use based on usewear analysis. After I had 
undertaken a whole tool extraction and counted the starch recovered, the 
sonic cleaned artefacts were given to Fullagar without any knowledge of my 
counts and the results compared.  That study found a strong correlation 
between high starch counts (an order of magnitude higher) than either 
unused flakes or the soils fromt the bag that contained the artefacts.  
That helped satisfy me that the starch residues were a) a real phenomena 
related to tool use (unless starch granules know which tools were used in 
the past and migrate towards them) and b) that cross-contamination from 
sediments does occur but will reflect the general counts of starch 
encountered in soils.  This of course raises further issues of depostional 
context and the definition of 'contamination' (which becomes a scale 
question). For example starch found on tools and in soils may relate to 
the history of site use, ie, that plants were processed here, contributing 
starch and tools into the archaeological record.

For a more recent blind test on starch granule identification see:

Mercader, J., et al. (2007) 4,300 year old chimpanzee sites and the 
origins of percussive stone technology. PNAS 104: 3043-3048.  

The starch identification test was undertaken by two analysts (Huw Barton 
and Robert Tyler) working independently, in fact the results were never 
disucussed by either party, so it gives a pretty good sense of what can be 
done and of the reliablity of single granule identifications to species or 
genus level (at least with this sample of nuts and yams)

In press currently I am publishing the results of small two-year study of 
starch preservation on tools (some left on the surface and some buried).  
It is a study I set up many years ago before I thought I would get into 
starch analysis in a major way, so many variables are not recorded such as 
soil Ph, etc. however, I found that starch (sweet potato) was preserved 
for two years on tools left on the surface and buried. this study will 
surface in a special volume being prepared in honour of Tom Loy.  I also 
noted that many starch granules were completely surrounded by very small 
particulate matter, which I believe (and see arguments in the Taphonomy 
chapter in Ancient Starch Research, pp. 84-85 on particulate organic 
matter) can effectively shield starch from microbial attack - if the 
microbes cannot reach the granule, it cannot be degraded, at least not by 
that mechanism.  I also think that once starch gets into a soil - of which 
there will be a great quantity - it will be under most threat of decay in 
the upper humic horizon, what escapes (and again see papers on the 
properties of Particulate Organic Matter in soils) is essentially locked 
up in the sediment, and variables such as depth of burial and pore size of 
sediments will also be important - compaction can be such that bacteria 
can no longer move.  The presence of water is also extremely important - 
as this is the vehicle for moving bacteria in soil.

Further good studies looking at starch in sediments are:

Balme, J and Beck, W. 2002. Starch and charocoal: useful measures of 
activity areas in archaeological rockshelters.  Journal of Archaeological 
Science 29: 157-66.

Atchison, J. and Fullagar, R. 1998. Starch residues on pounding implements 
from Jimnium rock-shelter. In Fullagar, R. ed. A Closer Look: recent 
Australian Studies of Stone tools, pp. 109-126. Sydney University 
ARchaeological Methods Series 6.  

Chapter 8 by Robin Torrence, 'Starch in Sediments' in the Ancient Starch 
Analysis book is also worth a look in the context of this debate and 
reviews some useful qualitative and quantitative experiements.

Recent work of my own has also looked at starch in sediments from a cave 
site in Borneo e.g. 

Barton, H. 2005. The Case for Rainforest Foragers: The Starch record at 
Niah Cave, Sarawak, Asian Perspectives 44(1): 56-72. (discusses the aroid, 
tuber starch and sago starch in cave sediments - people must have carried 
plant material into the site, its entrance is well above the forest floor 
and decaying underground tubers are unlikely to produce quantities of 
airborne starch -though a flour mill might.

Barton, H. and Paz, V. Subterranean Diets in the Tropical Rain Forests of 
Sarawak, Malaysia, in Denham et al. (eds). Rethinking Agriculture, pp. 50-
77. this paper further contextualises the results of the starch granule 
analysis by comparison with charred plant macro-remains from the site.  

There is every reason to expect starch in 'ancient' sediments is in 
fact 'ancient' starch.  While some people have raised the issue of 
possible post-depositional transport of modern starch into old sediments - 
it must occur to some degree - but such a mechanism must equally apply to 
all particulate matter of similar size, ie., pollen, phytoliths, and micro 
charcaol etc. It cannot all be contamination - or a lot of pollen analysts 
are going to be very unhappy people. 

As for dating, that would be great experiment to do, but I am not sure 
whether AMS could handle the one or even ten starch granules recovered 
from most sediment extractions - dating a tool residue with starch might 
be more likely to give a good result, even so, the quantites of organic 
matter are very small. 

I can also help out with starch mimics, if anyone wants to send me pics.  
Also I had thought of setting up a web site with a lot of that material, 
but wasn't sure there would be much interest-though this thread suggests 
otherwise.