[Gretl-devel] VECM experts ahoy!

I'm trying to upgrade gretl's testing for cointegration rank in a variety of cases. But I'm stuck on one point: understanding the relationship between Johansen's famous "five cases" (lucidly described by Jack in the Gretl User's Guide) and the five cases discussed by Pesaran, Shin and Smith (Journal of Econometrics, 2000), or "PSS" for short. Differences arise in relation to the "unrestricted constant" and "unrestricted trend" cases only. Let's forget the unrestricted trend, leaving the unrestricted constant. I think I understand Johansen on this (thanks, Jack!). If the constant, $\mu$, is restricted to the cointegration space its effect does not "leak out" into a trend, since $\mu$ is expressible as a linear combination of the columns of $\alpha$, and $\alpha_{\perp}'\mu = 0$. If the constant is not so restricted, it does leak out and generate a trend in the levels of the (possibly) cointegrated variables (call them Y). Now, as I understand it, PSS have a "Case II" which is identical to Johansen's restricted constant. Fine. But they also have a "Case III" which in some way resembles Johansen's unrestricted constant case except that it does _not_ generate a trend in Y. This I'm not getting. Can anyone help? I don't understand how you can "unrestrict" the constant and yet not have it leak out of the cointegration space into a trend in Y. Allin