The issue you raise, namely, the difficulty of getting an unequivocal structure ID of a "total unknown" in a complex mixture solely by chromatography-based mass spectrometry is far more difficult than most practicing scientists realize.
Even with accurate mass combined with MS^n, getting a structure with just chrom+MS alone is very challenging. For example, consider that a single elemental formula, C15 H22 O2, which has a nominal mass of 234 Da, has over 100 billion (yes, > 10^11) valid chemical structures.
I have performed mass spectrometry based structure elucidation in a number of roles, which have included working for a large pharmaceutical company supporting medicinal chemistry, as well as drug metabolism and pharmacokinetics. In addition I have also performed structure elucidation for metabolomics studies.
Of these cases, the metabolomics work presented the greatest challenge for the following reason. In my pharma role, I usually had some sort of information beyond just the mass spec data about the structure I was being asked to elucidate. For example, in the case of synthesis, the chemist knew what starting materials were present in the reaction, as well as the expected products. Often, they could produce different analogues.
In the case of drug metabolism, I always had the parent compound, and its fragmentation pattern combined with a knowledge of "typical" biotransformations facilitated ID of the metabolites.
Now, for metabolomics, the number of possible structures is far less constrained than in my previous roles. Having a comprehensive library of compounds will give you a good start.
You have searched a good collection of databases, but I recommend that you try ChemSpider (
http://www.chemspider.com/). It's free, and they currently index over 20 million compounds, including all of KEGG and HMDB just to name a few.
Recognize that many endogenous biochemicals can undergo the same transformations as xenobiotics (e.g. oxidations, sulfations, glucuronidations, phosphorylations, glutathione conjugations, etc), and that many of these conjugated forms will not be listed in databases (yet).
Here's the typical process I follow when trying to elucidate a "total" unknown.
1) Acquire accurate mass, MS^n data for at least n >=3.
2) Use accurate mass + isotopic information to reduce number of possible empirical formulas (Oliver Fiehn's "Seven Golden Rules" paper is a good reference).
3) Using the formula(s), search ChemSpider. Be aware that in LC/MS, not all compounds ionize via protonation. For example, carnitines and cholines, which contain a trimethyl-ammonium functionality, show up as the M+ species, not M+H like many others. Checking for Ring/Double/Triple bond equivalents is useful.
4) Frequently, you won't get any hits, so as one previous poster stated, look in the primary MS for common losses. For example glucuronides (-176), sulfate/phosphate (-80, but distinguishiable with accurate mass), acetic acid (-60), formic acid (-46), methanol (-32), etc all give clues to the structure of the compound. Also, sometimes, a match for the parent compound won't be found, but you will get a match for one of these decomposed species.
All of the vendors of accurate mass instrumentation have greatly oversold the capability of mass spectrometry for performing total unknown eludication. However, occasionally, you can identify a potential structure for an unknown, but, in the absence of some other information besides the mass spec data, in most cases, you'll be left with many possibilities.
Good Luck!
