Properly assembled organelle genomes are vital to fields from medicine to phylogenetics. New long-read sequencers provide opportunities for improving or even perfecting organelle assemblies, but little is known about the best way to combine long- and short-read assemblies to get the best results. We used the chloroplast genome as a test case, because it contains two large inverted repeats which make it difficult to assemble with only short-read data. Oxford Nanopore long-reads can span the whole chloroplast genome, but have a high error rate. We first developed a method for independently assessing genome assembly quality. We then investigated a range of assembly approaches using long-read only, short-read only and hybrid (i.e. combined long- and short-read) methods. In addition, we determined the minimum coverage of long- and short-reads necessary to achieve the best results. We show that hybrid assemblies with at least 20x coverage of long-reads and 40x coverage of short-reads produce the best results, with a single contig spanning the entire chloroplast genome with a very low error rate. Our results inform best-practice for assembling both animal and plant organelle genomes.