I have SRA IDs to download Salmonella sequences from the SRA NCBI web page. Then I need to assemble it. It is a valid a approach to do a De Novo assembly and then a Reference assembly with the output contigs/scaffolds of De Novo asembly. For example doing first a De Novo assembly with SPAdes and then mapping with BWA to my reference. My goal is to get a good assembly and then perform an annotation and in the end do a pangenomic analysis of the genomes. I am interested principally in the Newport serovar, so I am not sure if only doing the reference assembly with the genome of Salmonella enterica Typhimurium from NCBI or if in the end will be better to do De Novo assembly and the improve it with the reference.
If you can get a trusted Salmonella genome, I recommend you to use an assembler (such as Velvet, etc..) and then do a comparison between your assembly and the trusted genome with a program like Mauve
You will need to check how well the assembler work after selecting several k-mer for the assembling, since the choice of this kmer influences a lot the assembly itself
In the absence of a trusted genome to compare, chances to get a weird draft genome is high.