Hi Petar,
What is 'correct' will depend on what you want to do with the data after your mapping process. The alternative sequences (patches and haplotypes) are representations of the same region of the genome. Many people decide to work only with the primary assembly due to the redundancy of some of these sequences for processes like performing alignments. On the other hand, understanding the alternative sequences that have been found in this region of the genome may be important for your project.
Here are some brief definitions to help you make your decision:
Primary assembly
The underlying genome sequence, without alternative sequence included.
Haplotype (genome)
Known variations to the primary assembly, due to variability in the human genome sequence (eg. the highly variable MHC locus). These were included as part of the genome assembly when it was first produced.
Patch
New sequences that have been added to the genome assembly since its release. There are two types: fix and novel patches.
Finally, I've noticed that you are using BioMart for this mapping process, which is great. However, depending on the size of your dataset, you could also consider using the Variant Effect Predictor (VEP): http://www.ensembl.org/info/docs/tools/vep/index.html
BioMart is only useful for small and medium-sized datasets (approx. 500 variants per query), while the VEP can process millions of variants per query.
Best wishes
Ben
Ensembl Helpdesk