You can visually inspect the SV call regions in Savant which gives nice plots like these:

Loading a bed file with the low complexity and repeat regions of your genome of interest also helps with the visual inspection.

http://genomesavant.com/p/home/index

Reliable SV calling is still pretty much a research subject and you can expect a lot of false positive and false negative calls. Most people prefer to have false negatives so they are very strict about accepting SV calls and require evidence from multiple signals (depth, discordant pair, split-read / contig) and multiple tools to accept SV calls.

Some recent SV callers already use multiple signals for more reliable SV calling. Delly is an example that uses discordant pair and split-read https://github.com/tobiasrausch/delly

In the end you can design primers and sequence over the breakpoint(s) to be more sure about the SV's.

If you are talking about validating a larger region (>1000bp?) you could use Taqman Copy Number assays to validate the predicted copy-number (or manually designed qPCR to predict copy-number).

Also see: Confirming Structural Variants In The Genome