If they are true inserts/deletions then I would think they should be considered when defining haplotypes. If they are missing data then they should be coded as "?" and not considered.

The answer  depends on the quality of the alignment. If the quality is poor, it is likely that the indels are not real, and therefore should be excluded. If the indel frequency is very rare, I think it is likely to represent a sequencing artifact. If the allele frequencies of the the indel are intermediate, it is likely that the indel is real, and therefore should be included. Needless to say, that if you're analysing a coding region, you should not see indels that are not multiple of 3 (codons). I think in most software  the default is set to excluding sites with indels.

It is totally depends on your hypothesis of the research, actually what do you need?

There are various penalty methods as we all know. In the case of structural alignment (Various methods are there to evaluate like DALI, SSAP method and combinatorial extension method) and in phylogenetic analysis assessment of significance is required. 

So dear Pritesh the scoring score in between  your sequence to the existing sequences should be more that 99% than only you can think about your question. IF then you need to consider the whole thing.

all the best

I hope the question must have been solved. I am quite late here to discuss the question but I am curious to know how DnaSP handles the gaps/missing data to estimate haplotype diversity. The presence of a tandem repeat in a locus/gene of interest can generate gaps in the alignment and I assume the gaps generated here not due to artifact. By default, DnaSP excludes the gaps while estimating haplotype diversity, which in turn won’t be the accurate estimation of diversity. The other way to estimate nucleotide diversity of gaps/missing data is also available in DnaSP. I wonder, would I be wrong if I use the diversity estimated (form gaps/missing data) here is the diversity of the repeat region.