I agree that the Diamond Princess Cruise is probably the best population of tested individuals we have for guestimating infection rates, hospitalizations, and mortality. That said, in addition to being a small sample, there are a number of confounds with that data, most of which suggest the US will probably fare better than that particular population:

First, the average population on a cruise line is much older than a typical population (you mention this) and much older than the average US population. If you take age discrepancy into account and project over the US you end up with mortality rate well below 1% (the Ioannidis article from a couple of weeks ago estimated mortality between .05% and 1% based on that sample and he assumed some in critical condition would die after his article so he built that into his error rate at that time).

Second, even among the younger crowd on the cruiseline, it's not clear that they are representative of the general population in terms of overall health. My experiences on cruise ships is that there is a pretty substantial cohort of obese to extremely obese individuals on board (just check out the Lido deck during the crummy noon buffet and you'll see a lot of people that you don't see poolside or on the ropes course, if you know what I mean). I'm not saying that the cruisers are definitely less healthy than the average US person of similar age, but that is certainly possible and any variance will impact the accuracy of projections.

Lastly, the Diamond Princess cruise occured before many, if any, asymptomatic infections had occurred and been recovered from. These individuals build up anti-bodies and are unlikely to get infected a second time (not enough data to know if infection results in immunity, but that's typical for viral infections and so this assumption is not unreasonable). Prior asymptomatic (or minor symptom) infections will tend to push the CFR down over time.

I will say that the fact that the Diamond Princess passengers were not subject to an over-run healthcare system would cut in the other direction, but most factors with that data suggest the US would fare better than that cruise overall. It's kind of shocking that the Diamond Princess cruise remains one of our best data points even several weeks later (you'd think someone at the WHO or CDC would have tested an entire population in a town, city, or even a neighborhood to get better data...not to mention we're not testing anti-bodies at all, but for some reason we're in a data vacuum on real data but inundated with shoddy numbers using "confirmed cases" every day).

One potential take-away from the Diamond Princess cruise is that only 18.9% of the people exposed to the virus through re-circulated cruise-ship air actually contracted an infection and that's with an older population than in the US. Suppose 15% of the US would catch the virus without any preventative measures (that'd be 49M infected) and then .4% of those persons will die (that'd be 196K total deaths, which is about four times worse than a bad flu season). This would be without an over-run healthcare system (deaths could potentially increase a lot if that happens), but this estimate is also without taking any preventative measures (like social distancing, increased hand washing, and quarantined at-risk populations...all of which we are doing now).

My best guess using the Diamond Princess cruise data is that deaths in the US will probably exceed 50K but not 150K. Again, this is using the small sample data from the Diamond Princess cruise ship together with all of the preventative measures that are being deployed currently. I don't think anyone saying deaths in the US will reach millions has any credibility.