As Daniel mentioned, the deeper you sequence, the more representative of a population you will have in your data regardless of library preparation, especially with WGS metagenomics. There is a measure called rarefaction that tracks this in ecology (which is essentially what you're doing with metagenomics). Take a look at a few papers on metagenomics measures. A few other points as well: you probably wont be able to get coverage for de novo assembly (whole genome mapping) and wont be able to entirely separate out one organism vs another. Many metagenomics approaches rely on binning instead. There are some comprehensive reviews out there that you should take a look at before designing your experiment. A search for metagenomics review or primer will pull them up.

Yes it is possible that you will miss parts - this is because you're sampling by sequencing. In order to be sure you covered everything you would potentially have to sequence very deeply. Some organisms will be present at low levels, so how much sequencing would you need to do to a) detect them and then b) ensure they were fully covered? This would be a case of diminishing returns.

Whether this is an issue of library prep or not is almost immaterial.