[c-nsp] BGP Backdoor Links Problem

[Fossett, Jeff S]

Hi Team - figured most of you could provide a fix for the following scenario in your sleep, so I thought I'd reach out.
We have a Primary DataCenter and a DR Facility, attached via MPLS privately (using EIGRP across that link to peer the two DataCenters)... both also have an eBGP Peering Relationship with a Service Provider that is providing a private Cellular APN.   Though both connections are "Active" to the Service Provider Cellular PN, we want an "Active/Standby" type of routing scenario.  Our DR Center is not built up in a manner that adequately supports Active/Active from an Application/Middle Tier/Mainframe standpoint.
Scenario:  In our BGP peering relationship with the ISP (giving Cell Services) we are running eBGP.  Whenever routes are learned with eBGP they have a local cost of 20.  Since our "DataCenter Gateway" is getting the routes to the cell private networks from eBGP via "DR Gateway" they are always preferred over the Portland learned EIGRP external routes with a 170 cost.  When peering to the cell carriers is active in both the DataCenter and DR, the data sourced or destined for DR from the cell carrier networks will fail.
Problem:  We used the BGP Backdoor option on the DataCenter gateway to change the distance on the Cellular Cloud networks from 20 to 200 in order to allow our DataCenter gateway to prefer EIGRP routes within the DataCenter environment, and then redistribute into BGP.
But once we put this in, BGP no longer advertises any networks that are associated with the backdoor command.  Should we pursue some sort of Conditional Advertisement, or Synchronization/Redistribution scenario???
Any advice?  Thanks so much in advance for the help . . .