It depends if your gene has a pseudogene - in that case mapping to human genome is more correct, otherwise your all reads (for gene and pseudogene) will be mapped to your specified region. If your next step include variant detection, step2 will produce a lot of mistakes (false positives). It also depends how much pseudogene differ from gene - if the difference is small, also mapping to human genome can be wrong.
Best,
Agata
Option 1: whole genome mapping is obviously slower, but it is more accurate as the reads will map onto the most similar region it finds.
Option 2: a way faster, but may provide wrong results (i.e. reads of a sequence originated from an other part of the genome).
It really depends on how was designed the sample preparation step, but if you can afford the whole genome mapping in terms of duration, I would recommend this option strongly to avoid false positive findings.