It would be nice if Transmode spent a little more attention to the detail in
their data sheet and less on their art work. I can't discern what their CWDM
spacings are either. If you return to the site you'll note that the 1.25 and 2.5
Gxx numbers you cited were actually *bit rates*, not frequency separations on a
grid.
The ways in which grids are defined for DWDM and CWDM differ, with DWDM grids
typically defined on the basis of frequency spacings. E.g., spacings between the
center frequencies of lambdas at 200GHz, 100GHz and 50GHz proposed for the ITU
grids. The proposed CWDM Grids, on the other hand, will use *wavelength* spacings
(effectively the same thing after you do the arithmetic, it's just noted
differently).
See the following TIA (Telecommunications Industry Association) paper at the url
below. E.g., CWDM grid spacings may be at 20 nanometers (20 nm), whereas a DWDM
spacing might be 100 GHz.
http://ftp.tiaonline.org/fo-2/fo27/Public/2002-01-Kauai/FO27-2002-01-TD12.doc
I don't know how much this helps in the search for a mix and match wavelength
conversion assembly, but it at least lays out all known and proposed grid formats
so that we have a more factually-based starting point.
FAC