Alright, here's what I see as viable:
First and foremost, I'd aim for a large equipment order to backstop a lot of the expansion and to hopefully allow the NEC to get some new non-Acela equipment. The fact that the Regionals are all running with equipment ordered under Richard Nixon and delivered under Gerald Ford is getting a bit embarassing, and we're starting to run into practical capacity constraints that Amtrak is trying to "solve" with rising fares (amid a next-to-zero inflation economy, this is not going to be a long-term winner). Basically I'd look at a Viewliner-based body (which Amtrak is rumored to be looking at anyway) and aim for several hundred coaches alongside a batch of new cafes/diners and sleepers.
-Revive a "Broadway Limited", albeit sharing tracks with the Pennsylvanian (NYP-PGH) and Cap (PGH-CHI).
--New train would aim for an arrival into NYP around 0900. Resulting arrivals in WAS/PHL would be around 0730; departure from PGH would probably be around midnight eastbound.
-Run a section from the Capitol Limited to New York attached to the Pennsylvanian (the existing plan).
--Slide the Cap later in the day (have it as the "cleanup train") with a morning departure from PGH and a mid-afternoon arrival into WAS.
-Seriously look at a second Pennsylvanian with PA, aimed at a time distinct from the above.
--If NS wants an additional track for a train, I've no problem letting them have it...PROVIDED that we can extract some stack of slots in the process (e.g. non-expiring options for 5-6 r/t daily on the route).
-Second Lake Shore Limited. Time one LSL for an "early" arrival into New York (I'm thinking 0900-0930) which would likely translate into an early afternoon departure from Chicago (and thus only connect with regional hub trains) but connect with the Star, Crescent, and Meteor; and the other for something closer to the present arrangement.
-Daily Cardinal. Duh.
--I'd like to see a second train here, but there are complicated issues which make this less-than-likely.
With All Aboard Florida, I'd like to negotiate either some form of thru-ticketing or through cars or something like that. The best option I can come up with here would be to agree to buy perhaps 8 coaches and 8 sleepers for through-service (that would give you a coach and a sleeper on the Star and Meteor) and simply stick the cars onto whatever the next train out of JAX is (put in a large pad by all means but be willing to accept that the cars might go out on a 1400 train instead of a 1300 train so AAF doesn't have to hold their service for connecting equipment).
-I'd also negotiate to extend the Palmetto down to Miami via FEC's tracks, but dispatch it at something more like a freight train (e.g. depart JAX at about midnight and arrive Miami around 0700-0800...look at the heavyweight Havana Special for what I have in mind; hell, if you can get a ferry to through-ticket with it and establish a reliable connection, that's not a bad name).
My thinking is that the above should add no more than $10m to the direct losses of the system (and probably not even that...the "existing" trains would probably see their bottom lines improve and the added through-traffic on various routes should help as well).
Ok, onto the Midwest:
-Shift Amtrak to the Des Moines route for the most part. If a train gets left on the Burlington line (perhaps a stub train or split of some sort) that's fine, though it's an open question as to how much business a lone day train on this route with connections would have.
-Add a Denver Zephyr aimed at an evening departure from Chicago WB and a morning arrival into Chicago EB (which would connect legally with the second Lake Shore, Broadway, etc.).
--The result would probably be a mid-afternoon departure from Denver and a mid-late morning arrival into Denver. I'd probably adjust the existing CZ timing slightly to get a "daylight" CHI-OMA market in the deal (e.g. aim for an arrival into Omaha by 2130 at the latest). It might also be worth making the California Zephyr the "late" train while the Denver Zephyr gets the "earlier" slot and is aimed at CHI-OMA/OMA-DEN and CHI-DEN. Eastbound, the later train out of Denver would ideally leave Omaha no earlier than 0700.
-I'd seriously consider a second train into North Dakota if ND was willing to support service, aimed at a daylight timing west of MSP and overnight east of MSP.
--I'd pair this with the MSP-Duluth service and the CHI-MSP service that MN seems to want.
--This is all subject to MN and/or ND supporting such service(s), though I would think we might be able to get them to support some sort of "supplementary train" if PRIIA 209 can be worked around.
-The aforementioned beefing up of CHI-STL service (getting trains going faster, etc.) should proceed. I'd also add a St. Louis/Kansas City section on the Cardinal (as has been looked at).
Out West:
-Daily Sunset. Duh.
--Get the train re-routed into Phoenix. The current Maricopa stop just doesn't cut it and the tracks are still there; this is a good case for a TIGER grant.
-Get a day train LA-Tucson. This would require CA and/or AZ to get on board in some fashion (the shortest extension from LAX that would clear the 750-mile hurdle would be to El Paso, and I don't think there's enough traffic to justify that.
--I'd time this train so Phoenix-Tucson was a good commuter frequency (e.g. arrive PHX 0845, depart PHX 1730); given the size of this market and the distance involved, a non-stop or limited-stop run between the cities probably has the potential to fill the train[1]; if needed, I'd consider a pad on the way into Phoenix to try and ensure reliability. Being about a 10-hour trip overall, you'd get a pretty good timing for the LAX-PSP section that CA has wanted for a while as well.
--And of course, some of this would depend on the status of the AZ commuter rail project which has been slowly advancing. You've got increasing transit on both ends, but I know this is a big project...but having it there could whack a good bit of time off of the through-train.
-Expand service in Washington state. Subject to PRIIA 209, but Washington seems pretty workable.
-Desert Wind to SLC (with through cars). Net hit to direct losses of $10-15m, probably, but adds some good connections and probably helps with the Zephyr's bottom line (SLC-RNO is the "hole" on that route; added capacity CHI-SLC is probably fillable).
-I'd LIKE to see a Pioneer service of some sort, but I'm deeply reticent about both the added costs (I think Amtrak sandbagged their estimates a few years ago) and the prospect of having two "splits" on the Zephyr (I don't want a "City of Everywhere"; that gets operationally hard). I think that with only one train west of Denver, the Pioneer/Desert Wind question is an either/or proposition.
Backing up from the train projects I agree with pretty much all of what Nathanael has to say. In particular, the ASM needs to have some rules laid down about door opening (one door is fine at a small station but if there are more than X pax boarding a second door/door pair should be mandatory). Some of the points about cafe/diner service also come to mind (I know I'd happily take my steak to the SSL or cafe car if capacity was an issue).
I know I haven't hit everything here, but I think this is a decent starting point. I suspect that the western expansions (notwithstanding modest timing improvements and/or beefing up the fleet size a bit) will add to the overall losses while the eastern stuff will be mostly neutral on the direct losses side.
[1] An estimate from 2008 had something like 40,000 people commuting over 90 minutes in the region (
http://www.azcentral.com/news/articles/0107biz-tucsontophx0107.html); even assuming that only half of that is in a relevant market (e.g. people who live and work somewhere that shifting to transit would make sense), 1% market penetration plus through traffic to the LA area would pack a train pretty good (400 pax/day or 200/frequency); 2-3% penetration would overwhelm a single train (400-600 pax/train would probably effectively require the train to be about 8-9 cars long)