Is the E-term correction in PyAST and SLALIB the 'reference' implementation?

This is a question for @timj, @dsberry, and @scottransom - digging into the SLALIB source code (which I think AST relies on for the actual transformations?) it looks like the correction for the e-terms of aberration is linear in time. However, the Explanatory Supplement to the Astronomical Almanac suggests a 3rd order polynomial correction, which Kapteyn (and Astropy, in #551) implement. Should the version in SLALIB be updated, or is there a good reason for keeping it the way it is? (just trying to understand whether I should be using AST as a reference for conversions involving FK4 with the correction applied.