Mukaram Ali Khan, I'm sorry for not having noticed your question before now - nearly a year since you posted it.
As far as I'm concerned, I think that correlations between a particular construct (the main / target construct) and other constructs can provide evidence of discriminant validity if those correlations lie anywhere between minus .15 and plus .15. At a bit of a stretch, I'd even accept correlations lying between minus .20 and plus .20.