It may be that the EELVs effectively become exclusively USAF/NRO vehicles.
That won't happen
+1, and here is some of reasons why.
Although a little off topic below addresses why the Atlas V and F9 have such a cost difference. Plus for a NASA contract the payload integration, paperwork, and generally dealing with NASA costs for SpaceX is shown to be about $30M. ULAís price for this should be similar although its highly possible SpaceX underbid this effort due to lack of experience doing these tasks. This would make the price for the standard Atlas V <$135M before this extra services cost is added.
There are three items that are an economic thorn in the Atlas V price: vertical vehicle integration vs horizontal vehicle integration, in-house production vs outsourcing, and the costs of the 2nd stage.
First letís examine the costs of the two first stages. The engines costs of the RD-180 vs 9 M1Dís is nearly the same and the tank production is only slightly more for the Atlas than for the F9 making the cost difference due to the first stage minimal. In fact the Atlas V 1st stage cost could be less than F9ís due to such a cheap RD-180 cost. Here the in-house vs outsource cost difference has some effect and could make the Atlas V 1st stage easily cheaper than the F9 1st stage by bringing some of the more expensive 1st Stage components in-house.
Next letís examine the 2nd stages costs. The Centaurís tanks are cheaper to manufacture than F9ís, which is why they are still made the way they are and even future upgrades are still proposed for using stainless steel balloon tank structures. But this makes ground handling of these tanks slightly more difficult and expensive than a rigid tank so itís mostly a wash when it comes to costs. The avionics suite although very similar has a major difference that affects costs in a large way. The Centaur uses rad hard space qualified components throughout in a redundant system. F9 uses more modern off the shelf commercial components in a redundant system. The reliability difference between the two is not very much but the costs difference is significant. Most of the reliability is gained through the architecture not the reliability of the individual parts. This is true for operations taking place at LEO altitudes but not for GEO altitudes. Centaur unlike the F9 2nd stage actually performs operations/burns at GEO altitudes requiring the more radiation tolerant components. This gives an advantage to Atlas V/ Centaur for use in most BEO and some GEO missions over the F9. But for the rest of the payloads this double or triple the cost of the avionics is not warranted.
Also on the Centaur the extremely high cost of the RL-10 possibly now higher than $40M (in 2010 they were priced at $38M) each vs the MVAC at less than $2M shows up the major cost contributor to the Atlas V. Also the high outsourcing of other Centaur systems and components adds some cost increase over similar systems on the F9 2nd stage such as avionics and RCS. Putting the cost of the Centaur at >$50M each for a single engine Centaur. A DEC would cost >$90M. The cost of the F9 2nd stage is probably <$15M. A $35-75M difference. ULA really needs to replace the RL-10 with a cheaper modern engine with somewhat same form-fit-function but costing <$5M each. One additional item about Centaur and that is there is a structural limit for payload size of about 9,000kg. It would require a thicker stainless steel tank and other structure improvements to compete for the heavy LEO payloads. The DEC would be made for heavier LEO payloads leaving the single engine Centaur for all other payloads below 9,000kg. There are some costs increases associated with this as well as a performance penalty due to increased dry weight, but performance loss for LEO is offset by the increased performance when using 2 engines (less gravity losses).
Last is the vertical vehicle integration vs horizontal vehicle integration. It has been argued back and forth which method is generally better overall and which is generally cheaper overall. If most of your launches are commercial satellites then horizontal will be much cheaper but still not half the cost, somewhere around a 20% cost savings. But if most of your launches are government, horizontal could cost more than vertical due to the payload handling requirements, specialized jigs, etc.
ULA could become very competitive price wise for any government contracts and even commercial if they do a few cost reduction items: replace the RL-10 on the ACES replacing the Centaur (it has cost savings all over the place making its costs about even with that of the F9 2nd stage) and produce parts/components in-house that no longer have qualified sources. Doing this could get the price of an Atlas V down to $80M or $75M making it price competitive with Proton and depending on what the customer wants competitive with F9.
ULA payloads most at risk being lost to SpaceX are LEO payloads, even then SpaceX would probably only pick up about 50% of those in order to maintain space access assurance because most of the LEO launches are out of VAFB whereas the GEO, MEO, and BEO are out of the Cape with only a very few LEO launches.