As I understand it (from Dynamics text books, not from STAAD), the process is the following:
1) Sum up the modal response for the known / solved modes.
2) If the left over modes can be assumed to be "rigid" then you can combine their response together into an assumed rigid response mode. You know the mass distribution of this mode based on whatever mass is left over when you subtract the solved modal response (sum of mode shape * modal participation factor) from the unity vector.
3) You know the acceleration of this mode is the rigid response acceleration of your spectra. With all of this known, you can use your stiffness matrix to back out a residual model shape.
Be cautious though. There are a lot of ways this calculation could go wrong. If you don't have the ability to independently confirm the results then I'm not sure I would trust STAAD to get it right. Though my experience with STAAD is mostly from the old REI days (10~15 years ago). Perhaps their QA/QC has improved since they were acquired by Bentley.
Disclaimer: I work for one of STAAD's competitors. Though most of my skepticism is based on my time with another engineering firm where I worked on my firms NQA testing / validation for STAAD, SAP, and RISA. That time frame means that I was working with some of the last versions of STAAD-III and maybe the early versions of STAAD-Pro. That could have just been a difficult transition for them.... But, it drove me bonkers at the time and instilled in me a deep seated skepticism of computer output in general and STAAD in particular.