Is it a good practice to point out errors (politely) in a previously published paper (from which I have borrowed idea) if the reviewers explicitly asks to do so in a review response. I mean the reviewer has asked "why you have made this strong assumption, this was something weak in the published paper". In this situation, I think I have no option other than pointing out the errors for which I have enough mathematical evidence, I even asked the author of the published paper about that, he also agreed about that.
If you are building on the work of a previous paper, but the older paper contains a material error, you should not be waiting for a referee to ask about the discrepancy. A discussion of the error should be included in the manuscript from the start. It could run something like
The outline of our calculation is similar to that in . However, it should be noted that  actually contains an error in the calculation of the r coefficient....
If your paper disagrees with the earlier literature that it relies on, readers trying to understand the details of your work are likely to be confused. If a reader notices the discrepancy, they are probably most likely to think that you have made an error. The referee who looked at your manuscript quite likely brought the discrepancy to your attention, and the right solution is the correct the record before you move on to your own new results.