I had it convert to straight strings, edited it manually until it was right, and ended up with (including a separate non-capturing move) vDvmpafcaKvmpafcabW, which works perfectly.
The vD and vmpafcabW seem to be extra compared to the bracket notation (where a also means all but back to where you came from). I am not sure K/Q after D is not recoginized as a case where it should expand the first leg into K steps, in the pre-processor for bracket notation. This is a combination that only makes sense when the second leg can freely change direction; for fixed direction you could use F/W/B/R. I am sure the bracket notation would work for those (e.g. the Snake Tongue). It could also be that the a in the bracket notation is not removed. I will have a look at it.
[Edit] OK, [vcD-aK] now works. The problem was two-fold. For one, the prefixes of the first leg (vc here) were kept in front, because they were expected to indicate the initial direction. But c is a mode modifier, that should go with the last leg of the expanded D. Only the v should stay in front.
The second problem was that the atom it picked for the XBetza description was W for an orthogonal leap like D. And although it correctly interpreted the a in aK as 'all directions', 'all' means something different for W than for K. It should have recognized K as a continuation to overrule the W or F choice. This will still give problems with complex directional specifications, though. But fortunately not here, as vK means the same as vW. But had the first leg been fA or even fsD, the directions would have meant something different on K. I haven't solved that yet. Perhaps this has to be saved for when the bracket notation can really be parsed directly, rather than through pre-processing.
The vD and vmpafcabW seem to be extra compared to the bracket notation (where a also means all but back to where you came from). I am not sure K/Q after D is not recoginized as a case where it should expand the first leg into K steps, in the pre-processor for bracket notation. This is a combination that only makes sense when the second leg can freely change direction; for fixed direction you could use F/W/B/R. I am sure the bracket notation would work for those (e.g. the Snake Tongue). It could also be that the a in the bracket notation is not removed. I will have a look at it.
[Edit] OK, [vcD-aK] now works. The problem was two-fold. For one, the prefixes of the first leg (vc here) were kept in front, because they were expected to indicate the initial direction. But c is a mode modifier, that should go with the last leg of the expanded D. Only the v should stay in front.
The second problem was that the atom it picked for the XBetza description was W for an orthogonal leap like D. And although it correctly interpreted the a in aK as 'all directions', 'all' means something different for W than for K. It should have recognized K as a continuation to overrule the W or F choice. This will still give problems with complex directional specifications, though. But fortunately not here, as vK means the same as vW. But had the first leg been fA or even fsD, the directions would have meant something different on K. I haven't solved that yet. Perhaps this has to be saved for when the bracket notation can really be parsed directly, rather than through pre-processing.