Target capture necessarily requires subjecting your libraries to a bottleneck, wherein target molecules are captured and enriched, and non-target molecules are therefore removed. To have sufficient unique molecules for good sequencing coverage of your targets, successful captures DEPEND on the input of sufficiently complex libraries. For this reason, for libraries with a significant non-target component (e.g., ancient, forensic, or environmental samples), and especially for WGE captures with a very large full nuclear genome target size, we strongly recommend maximizing the target component in each capture by using as much input library as possible (up to 2 µg+), and consider two rounds of capture for higher percentage of reads on-target.
For best results, it is recommended that only amplified (non-PCR-free) NGS libraries are used for target capture. This provides multiple copies of each starting template molecule, increasing the chance of each individual molecule getting enriched. However if you need more starting material to reach the recommended amount, it is generally preferable to generate more library from fresh genomic DNA or a new batch of indexed library, rather than through extra amplification. This is because while some amplification is good, over-amplification risks reducing the observable complexity of your libraries through the uneven action of PCR bias, as some molecules will become relatively more abundant while others become rare. This is also true for manipulating your libraries after capture: amplify your post-capture libraries the minimum number of cycles necessary to reach the molarity required by your sequencing facility.