"I presume you know that 16S isn't subject to horizontal transfer. So your point is?"
It isn't?
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=14645285&dopt=Abstract
http://www.ttaxus.com/files/badger2005-hyphophylo.pdf
Likewise, even though 16S is sometimes useful as a tagging mechanism, it hasn't even shown to be phylogenically useful above the genus level in many eukaryotes:
http://www.ttaxus.com/files/badger2005-hyphophylo.pdf
So, it's not even useful for declaring universal common ancestry, because even with the gigantic limitations of the methods, it STILL doesn't produce a workable tree when you go beyond unicellular organisms.
"But we know it happens. We observe it in modern bacteria."
I agreed to this point. The problem is that the existence of horizontal transfer makes it impossible to differentiate gene features that are based on horizontal transfer and gene features that are based on unique origin. You can't tell the difference simply from looking at the features.
This is the problem of modern phylogenics -- the phylogenic methods all assume that universal common ancestry is true, so they don't have cases for saying "this isn't a case of universal common ancestry". I can build a best-fit tree for ANY set of data. This doesn't prove that a tree exists. For one to be able to determine _whether_ a tree exists, I would have to have criteria for which I would say "no tree exists". However, now, with horizontal transfer, you wipe away even common-sense non-trees by simply filling in the problem areas with horizontal transfer.
It is all based on the _assumption_ of universal common ancestry. The only way that it isn't an assumption is with a workable model of abiogenesis, which does not exist.
For one to be able to determine _whether_ a tree exists, I would have to have criteria for which I would say "no tree exists". However, now, with horizontal transfer, you wipe away even common-sense non-trees by simply filling in the problem areas with horizontal transfer.
You can do a rigorous mathematical analysis of the data, and make a decision on whether a tree exists; such an analysis will be complicated but not rendered invalid by HGT. HGT only gets to be a serious problem when the rate of transfer becomes comparable to the rate of genetic divergence. This isn't true in higher organisms, and it doesn't seem to be true in prokaryotes; after all, even in the second paper you cited, by including more genes (better statistics) in the analysis, they came up with what they claim is a correct taxonomy.
Ultimately, we'll be comparing genomes with genomes, not genes with genes. At that stage, it is possible that HGT will make the analysis of some branches impossible, but so far the indications are that it will not. And in the end, if HGT is such that we can at most trace ancestry to a some finite population of gene-transferring protists, that won't be the body-blow to evolution you hope. The LUCA is more an amusing corollary to evolution than a central dogma.