I think you're confused with what we call the "reference image", or sometimes the "master image". You are supposed to pick tie points between an orthorectified reference image (the Master ortho-image), and a raw slave image, in order to create GCP for the raw slave image. From your question, I am assuming that you picked tie points between two raw images.. hence the missing map information. This procedure is of course wrong, because you're not using any reference ortho-image.
If you do not have any reference ortho-image, we advise to create a shading version of the DEM, and use this synthetic image as the first ortho-master. Then, the first step would be to pick tie points between a first raw image and the shaded DEM, and convert them into GCP. If the DEM resolution is close to that of the image, or better, try optimizing the GCP produced. If the DEM resolution is much lower than that of the image, GCP optimization might fail, and in this case, GCP optimization is sueless and tie points should be selected very carefully.
Then you ortho-rectify your first image according to the first set of GCP, and the first ortho-image produced will become your new reference for your second raw image. Then, from now on, the first ortho-image is your reference for all subsequent processes, and GCP optimization should always work well.
Hope this help,