It is my belief that F-W### is the registration series for what in the USA would be described as 'experimental' aircraft (not necessarily small and homebuilt - I photographed an A.400 carrying the registration mark F-WWOW) and that if and when they satisfy the criteria of the CDN/CDNR or CNRA registers, they migrate to that usually, it seems, carrying the last three letters of their 'experimental' registration mark. So theoretically, if an aircraft carried the 'experimental' registration mark F-WAAA one would expect it subsequently to become F-B/G/HAAA (on the CDN/CDNR register) or F-PAAA (on the CNRA register). But that's not necessarily the case. Some aircraft seem to have carried their F-W### registration marks for ages and never migrated to a F-B/G/H/P### mark. F-WCDG is an example of this - F-BCDG was a M.S.500. Thus it leaves me wondering whether the DGAC reserve a CDN/CDNR/CNRA registration mark for an aircraft and then require it to carry the same mark but with the initial B/G/H/P replaced with a W during its 'experimental' period because, that way, there a better chance of being able to carry the last three letters of its registration mark from F-W### to F-B/G/H/P###, knowing that the latter has not already been allocated, rather than allocating a F-W### registration mark sequentially, only to find that when the aircraft migrates to the CCDN/CDNR/CNRA register it has to take up a mark with four different letter anteceding the national letter. And if the aircraft doesn't migrate, perhaps then the DGAC issues the correlating mark to another aircraft (as was the case with F-WCDG and F-BCDG). Anyhow, as I've said, that's my belief but I'll try to make enquiries of some who are more likely to know than believe!
P.s. please don't ask me what happens when an aircraft moved from 'experimental' to the French ULM register, the marks for which comprise the number of the département in which it is registered (e.g. 79 for Deux Sèvres) plus two or three letters (e.g. 79-AA).