I'm not sure if there's any formally accepted method for doing this, but most of the work I've read on this subject uses a consensus taxonomy of essential genes to make a determination of the phylogeny of uncultured genomes. I've attached a paper that uses this approach, but I'm sure there are newer ones around. Using this technique you end up with a phylogenetic tree from which you infer the nearest relative - there's no classified/unclassified state (which is a bit of an artificial delineation to begin with).

In general I think Candidate status is something that becomes established by consensus over time. Candidate phylum (like TG-1 used to be) were originally called such because there was extensive evidence that they existed (16S rRNA sequences, FISH images etc), being found in a range of independent studies around the world over many years, they just lacked a typestrain to 'prove' they were real. It's probably not something that would be reported in a stand-alone manuscript or report, but something that builds up over time and may be formalised in a literature review.

EDIT: Sorry, I meant a concatenated protein tree from essential genes, not a consensus taxonomy. Consensus taxonomy works too (i.e. BLAST each gene then work out the common ancestor for all sequences) but the concatenated tree approach gives you a better feel for what you're really working with.

Article Genome sequences of rare, uncultured bacteria obtained by di...

Anyone can deposit sequences in the NCBI database. The sequences are not checked by a third party. Thus I recommend to limit the search to type strains. These sequences are assingend to species descriptions and should be correct.

Nowadays the bacteriological code does not accept new species based on single cell genomics, genomes from environmental samples,... only if they can be cultured.

The 'Candidatus' status for all this bugs has become a non-official standard. You have a note about it in LPSN

http://www.bacterio.net/-candidatus.html

In the paper below you can see an example of a new Candidatus species

Article Single-cell and metagenomic analyses indicate a fermentative...

For the moment, for a novel species you should use

Actually 16S itself was discovered to use for identifying microbial phylogeny including novel taxa definitions and 97% was intended to represent species, but is should be supported by phenotypic data if you want to publish a VALID species. If not, and this is a case here (as I suppose) you can use just 16S. Moreover the whole species description is not so solid in microbiology since in the real life it could be a kind of continuum without species and possibly even genera...

Well, there is certain differences between "Genus species" and 'Candidatus Genus species'.

The style "Genus species" is applied to not validly published species or taxa. Despite of being tentatively published as a valid description, if under the bacteriological code it does not comply with the minimum requirements, the proposed name should be referred in posterior publications between quotation marks but still in italics. It indicates a non-standing name under the bacteriological code.

Using the form 'Candidatus Genus species' (abbreviated 'Ca. Genus species') in a paper, clearly indicates that you know that the description does not follow the rules of the bacteriological code. The ICSP encourages the use of the 'Candidatus' concept whenever an organism is well-characterised, but not yet culturable (see paper below). It means that having just a set of 16S sequences from your community profiling is not enough. Frequently, nowadays, this is achieved by genomic reconstruction of your bug from highly enriched samples, single cell genomics,.. Format: single quotation marks and only candidatus in italics.

Finally, the term candidate division or candidate phylum, is restricted to high rank taxa for which you don't need to have exhaustive information. This candidate divisions are usually determined by the presence of basal divergent branches in phylogenetic trees constructed with data from high throughput sequencing.

Article Report of the ad hoc committee for the re-evaluation of the ...

@ Daniel, I don`t understand what exactly you mean by "On the other side it is a common misconception that 97% OTUs are species". 16S or 97%? I`m trying to tell you I suppose the species concept is much more important for humans (microbiologists) than prokaryotes themselves. However the current tradition/rules/fashion, whatewer, forces us to make several definitions and we should follow them at least for valid species description to keep some kind of order. However, the rules should not be so strict for OTUs predictions and these rules should be uniform, understandable and usable in different fields, eg. for analysis of 16S pyrosequences AND complete metagenomes AND phylogeny of valid species and so on. We can use real DNA-DNA hybridization, but it is extremely time and resource-consuming (and innacurate) and could not be used in any metagenomic studies. We can try to use ANI, but it is possible if we have complete genomes or almost completed drafts. We can use 23S but again it could be found in complete genomes only (with few exceptions), we can use chemotaxonomy, but it is a kind of mess. Thus, historically  16S was a chosen one :) due to amount of 16S sequences and comparative data. I could not found the paper you provided, I suppose it`s a bit old-fashioned. I remember there was a paper of Stackebrandt and Ebers describing 98.6% of different species border, however I suppose 97% is a more solid and common.

1) As I know, bacterial species are assigned for a group  of strains/isolates with common genetic, physiological and morphological properties. To my opinion, again, it is difficult (impossible) to create a new taxon for an unique single isolate, which is similar to unique genome obtained from shotgun metagenomic assembly . 

 2) The species-level deconvolution of metagenome assemblies has a certain level of probability/uncertainty, thus, it will be difficult to obtain similar result in a number of shotgun metagenomic data from different experiments (Burton, J. N.; Liachko, I.; Dunham, M. J.; Shendure, J. (2014). "Species-Level Deconvolution of Metagenome Assemblies with Hi-C-Based Contact Probability Maps". G3: Genes, Genomes, Genetics.)

You can find the answer to your question in this paper:

Murray RG, Stackebrandt E. (1995) Taxonomic note: implementation of the provisional status Candidatus for incompletely described procaryotes. Int J Syst Bacteriol. 45:186-7.

http://www.ncbi.nlm.nih.gov/pubmed/7857801.

I think this paper might be of help, it is the most recent effort on the classification of uncultured bacteria and archaea:

http://www.nature.com/nrmicro/journal/v12/n9/abs/nrmicro3330.html

Yarza P, Yilmaz P, Pruesse E, Glockner FO, Ludwig W, Schleifer K-H, Whitman WB, Euzeby J, Amann R, Rossello-Mora R: Uniting the classification of cultured and uncultured bacteria and archaea using 16S rRNA gene sequences. Nat Rev Microbiol 2014, 12:635-645.