The statement is not 100% correct. Ordinary mass spectrometry will not give you exact masses to 5 or 6 decimal places, say, m/z 415.14 for a singly charged molecular ion. Now there will be hundreds of compounds with this mass. On the other hand HRMS gives you a accurate and precise mass say 415.14509. This detailed number drastically reduces the number of matches in a spectral library. Not only that, the isotopic distribution pattern and their masses reduces the choice to one compound. Hence the claim that HRMS will not need a reference.
Like any claim, there are always exceptions. Structural isomers can be a problem. If we have isotopologues (isotopic isomers), even HRMS will fail. Say you have two analytes but both are acetonitriles. Say one is
$^{13}$CH$_3$CN, and the other analyte is CH$_3$$^{13}$CN, HRMS will give the identical molecular mass to five decimal places. One might look up their fragmentation pattern but MS is not that great at low masses (high noise). Here you would need another spectroscopic technique such as microwave rotational spectroscopy. These techniques are used an in emerging field called position specific isotopic analysis.