Hi Quetzal,
The contamination problems (plural as there are many different problems) not only applies to old or ancient DNA but to modern samples as well. However, the low copy and degraded state of ancient DNA exacerbates the problem. Iwill try below to show why human or primate ancient DNA is particularly tricky relative to say work with mammoths.
Problems:
1. There is more DNA in a few cells of skin you shed than there probably is in most ancient human fossils. What this means is by the time you get the fossil from somebody who will allow you to whack it to pieces to extract DNA from it, lots of people have touched it with bare hands..not to mention those who did the excavation. If you cannot sample from an internal part of the sample i.e. taking a bone core then you will have a major contamination source. As Cooper in the Nature commentary states and has been demonstrated by others over the years, soaking in acid, bleach, UV light does not necessarily get rid of all contamination.
2. PCR primers that work for humans usually work for their relatives or near relatives. If I design PCR primers for mtDNA of Homo sapiens..they will work for Cro-Magnon as well. Contrast this to mammoths. I can design (and have frequently) primers that are specific for elephants. You can spike a reaction with human DNA and still never amplify it..so my Russian curator friends can touch the bones all they want and it won't cause me any problems.
3. Mungo Lake suffered from a different problem which plagues ancient DNA as well which are nuclear insertions of mtDNA or NUMTS. These are copies of mtDNA that have wandered into the nucleus and are now pseudogenes...there are hundreds if not thousands in the human genome...they often look very much like human mtDNA for example but often show differences from most bona fide mtDNA...they can be so divergent that they fall outside of the genetic variation of mtDNA for the human species i.e. like the neandertal type specimen sequence. That is why Krings et al. spent so much time and experimentation to support that the neanderal sequences was not an insert. The Mungo Lake sequence phylogenetically clustered with a known human mtDNA NUMT which is a big heads up that you have contamination...the most famous example is the claim that dinosaur DNA had been sequenced and it turned out to be a human nuclear insert sequence..doh!
With the above in mind...what do you do with ancient human samples? If the samples is contaminated your negative controls could still look good since your reagents are presumably not contaminated. You extract from a Cro-Magnon or a neandertal and get a sequence that looks like an average Joe mtDNA sequence. You send it to another lab to replicate the result (and since they will have the same contaminated sample) they may reproduce you contamination result...you can test multiple different samples and see if the results are consistent i.e. is the next Cro-Magnon sequence from another location similar or radically different? But if it comes back as a different common human mtDNA haplotype..what then? If it is completely different...which sequences is correct? Is one a contaminant? Are both?
Thus, you end up only accepting sequences that look very different from modern human mtDNA and that can be confirmed as non-nuclear in origin...and I think it is pretty pointless to do that much work if you cannot accept basically an entire array of data as valid because of contamination fears. Of course accepted neandertal sequences will be different from modern human since if a neandertal sequence with modern human sequences is found it will be dimissed as contamination!
Contamination, cross contamination, NUMTS all plague modern DNA studies as well...but a modern sample usually has so much DNA in it that you would really have to contaminate it pretty badly. Numts are a bigger problem which the scientific community mostly ignores rather than addresses...for my part, I have switched to nuclear DNA markers for mammoth analysis since my work with elephants showed they are so loaded with NUMTS that it is almost impossible to figure out what the bona fide mtDNA sequence is for the regions of the mitochondria that would interest me...of course, since so few mammoths yield nuclear DNA it has slowed my research down..but I rather deal with that then publish a completely screwed data set...and besides..its not like there is a big community of people working on mammoths...so my competition tends to be minimal compared to my more medically oriented work..so I can afford to be more plodding in my progress
...how is that for an excuse for being lazy