Any launch vehicle would. I have no doubt Griffin may retain his SDI roots. That doesn’t mean they transferred to SpaceX, much less Starlink.
The most we can say is SpaceX’s founders and allies were influenced by SDI. And that influence peaked in the 2000s, when SpaceX was dependent on grants versus launch contracts.
> Starship seems optimized for mass to LEO rather than Mars
That’s where the money is.
Given Starship’s flexibility, concluding much about a secret purpose is untenable. (I agree it would seem to need sizing up for Mars. But if in-orbit propellant transfer really nails, LEO and Mars are practically neighbours. (9.4 km/s to LEO, another 3 to 4 to Mars. More to the Moon.)
Who is related to Starlink in what respect?
The link between Starlink and SDI is about as close as it is to anything space related the U.S. government has done. It’s equally correct to say Starlink has its roots in space planes. Like, sure, I can draw tenuous links between the two. But that’s like saying Juicero was a military project because Silicon Valley was seeded by the DoD.
Not even remotely true. You can broadly classify SDI proposals into two categories; ones that require scientific breakthroughs, and ones that are really big engineering projects. The problem with scientific breakthroughs is you can't reasonably predict when and how they're going to happen, so these aren't serious proposals.
The other sort, the large engineering problems, require tens of thousands of satellites (e.g. interceptors). An oldspace style rocket with a few launches a year will never move the needle on such a project, it needs a rocket like Falcon 9 or Starship. Michael Griffin's background is in this category of SDI proposal, there can be no doubt that he knew how many satellites it will take and that doing so without cheap reusable rockets would never work.
Rockets like Vulcan are similarly capable for SDI purposes as the Falcon 1 was.
I'm open to being corrected by anyone else who was involved with SpaceX in its early days. But the SDI connection to Falcon reusable, much less Starlink, is an expert exercise in retconning.
Now, you said that even Atlas rockets developed after 2003 are technically advancing an SDI agenda. I don't know which Atlas rocket you're talking about (Atlas V Heavy development?), but it definitely isn't true. There was never any pretext of the Atlas rocket family being a step towards reusability, and therefore it has nothing to do with any serious SDI proposal.