This is well-written and analyzed, with one important caveat:

In the illustration of "Delaying Social Security — Single", the assertion is made that the retiree payments after year 1 are all 8% higher if he waits a year. This ignores the one cost-of-living increase that apply to all years for the case where he chooses to retire at 66. If the COLA is 2.5% (which is a reasonable long-term assumption), then there is only a 5.5% difference in payments starting in year 2.

Once this is taken into account, the entire NPV argument goes away altogether.

Other concerns can now be given their full weight. For example, should we really be encouraging clients to spend down their other savings in order to delay starting social security? Doing so shortens their investment horizon, and exposes them to more market volatility than they otherwise would have.