C-CDA to FHIR: copy section.text to section.text; copy entry.text to resource.text. Since entry.text is usually not populated, it may be necessary to develop or re-use rules to generate text from structured data, scoped to the resource rather than the section.

FHIR to C-CDA: copy section.text to section.text. Entry.text is usually not populated.

Strongly disagree with the statements about entry.text not being populated. This has been recommended since C-CDA 1.1 and more clearly in V1 of C-CDA 2.1: 

The C-CDA R1.1 release recommended that clinical statements include a link between the narrative (section.text) and coded clinical data (entry). Rather than repeat these constraints in every applicable entry, SDWG agreed in R2.0 to apply the following constraint to all entry templates, unless explicitly prohibited.

SHOULD contain zero or one [0..1] text (CONF:XXXX).

1. The text, if present, SHOULD contain zero or one [0..1] reference/@value (CONF: XXXX).
   1. This reference/@value SHALL begin with a '#' and SHALL point to its corresponding narrative (using the approach defined in CDA R2.0, section 4.3.5.1) (CONF: XXXX).

This is repeated in the companion guide:

Best practice for CDA creation is to represent all human readable text in the section, then reference the text from the discrete entries that represent the human readable information as machine processable data. To include narrative text linking, the text element of the primary (outer-most) act in an entry should point, by reference, to the portion of the narrative text corresponding to the meaning of the entire clinical statement expressed in the discrete entry.

 
So while it is certainly not required, this mapping should not be making the statement that it is "usually not populated" nor should it be recommending that, when converting FHIR to C-CDA that this is usually not populated.