I'm not dogmatic about taking it early or late. Frankly, based on what's knowable, it's about as close to a crap shoot as you can get. And the differences are generally pretty small. What people say about the differences in how they feel about blowing through SS dough vs pulling money out of the portfolio is as real as any math.
I did run a few scenarios on i-orp. I selected that tool because it's quick to run a scenario like this, and it doesn't ignore important things. What I did was do successive runs, switching between taking SS at 62 vs 70. First, the assumptions: AA of 60% equities that
returned (in most runs) 5% over inflation, but I ran a few with 7%. These are much lower than the historical CAGR for equities, but supposedly in line with what the experts are saying nowadays. CPI Inflation of 2%, spending inflation of 4% (those are defaults on the site). I varied the dirt nap through 87, 92, and 97, because obviously, the longer you live the more waiting makes sense. But really, the 97's would be an unlikely event for most of us, especially the males. But it's in there because I wanted to show the "take it late" option "winning", given a long enough life.
For the simple scenario proposed, the $1M IRA, presuming SS will continue to be fully funded, 60% stock allocation growing at 5% real, your annual spend could be between 1.4% (87), 3.0% (92) and 4.5% (97) higher by waiting until 70 to claim.
A similar scenario, $1M in Roth, presuming SS will not continue to be fully funded, 60% stock allocation growing at 7% real, your annual spend could be between 1.7% (87) lower, 0.5% (92) lower and 0.4% (97) higher by waiting until 70 to claim.
Yes, you can pick apart the second scenario I used. I simply picked one that I figured might be more favorable to taking it now, and it was. I didn't have to reduce the realism by turning up the equity allocation or anything. Just a slightly higher return and the default on the i-orp site is to presume SS will not continue past 2034 with full benefits. Still squarely in the realm of reality for some people. Anyway, without much effort I found a scenario where the "taking it late is always better" dogma can be proven false based on the math. Any scenario where anything close to the historical CAGR for equities is entered will give you a "take it now" answer as well.
As to the original scenario, there is not just one answer because there are too many assumptions that have been left out.