Do we have dromaeosaurid evolution backwards?

The basic evolution of Maniraptora has seemed pretty well established in the past decade, thanks to TWG papers describing Sinovenator, Mei, Mahakala, Xiaotingia and such.  The basal paravian was a little bird-like taxon, like Microraptor on the dromaeosaurid end or Jinfengopteryx on the troodontid end, with genera like Rahonavis and Anchiornis breaking the boundaries even more.  Going further towards the base of Maniraptora, the cranial similarities between scansoriopterygids and basal oviraptorosaurs have suggested a short-snouted herbivorous ancestor, while Shuvuuia and Pelecanimimus have similar skulls that suggest the first maniraptoriform was not a macropredator.  Large, more obviously carnivorous taxa like eudromaeosaurs are seen as reversals to a more traditional theropod lifestyle.  It's a nice story and may be right, but what if it's wrong?

Dromaeosaurid morphology forms a continuum from the extreme of Achillobator with its deep snout, mesially serrated teeth, low DSDI, rather short coracoid, posteriorly facing glenoid, relatively short arms, deep brevis fossa, large anterior pubic boot and proximally placed obturator process, though Deinonychus, Velociraptor, Bambiraptor, Sinornithosaurus, Microraptor, Buitreraptor/Unenlagia and ending at Rahonavis.  Even if the unenlagiines are avialans though, dromaeosaurids need a lot of reversals no matter which direction evolution went.  Note that stratigraphy doesn't strongly support either option.  We have Utahraptor from the Barremian which is very similar to Achillobator, and dromaeosaurid-like teeth with mesial serrations resembling both dromaeosaurines and velociraptorines in the Late Jurassic.  Then again, there's the microraptorian-like Graciliraptor and Shanag which also lived early, and the possible microraptorian "Paleopteryx" from the Morrison.

There's also a possible transitional form between basal coelurosaurs and dromaeosaurids- Ornitholestes.  Like dromaeosaurids, Ornitholestes has a third premaxillary tooth much smaller than the first two, short cervical vertebrae with tall neural spines, prominent anterior cervical epipophyses, a crest-like ventral tuberosity on the humerus, an enlarged second pedal ungual and a transversely expanded metatarsal IV.  It's also similar to paravians in the elongate distal caudals and bifurcated chevrons.  Deriving dromaeosaurids from something of Ornitholestes-grade would explain why they are almost unique among derived maniraptorans in having prefrontals, which unlike the dental characters of eudromaeosaurs, are not plausibly due to macropredatory habits.  In this scenario, microraptorians would be convergent with birds in their aerial characters.

Is there any other evidence for this idea?  Don't troodontids show the same pattern as dromaeosaurids, going from Jinfengopteryx/Anchiornis to Mei/Sinovenator to Sinornithoides/Byronosaurus to Troodon?  Maybe not.  Jinfengopteryx and Anchiornis can switch to Avialae easily, and the same may be true for Mei and Sinovenator.  They're often avialans in the in progress Lori matrix, even with Xu et al.'s troodontid characters.  Lori itself emerged sister to Sinornithoides in Hartman et al.'s SVP poster and is Jurassic in age, with serrated teeth.  Maybe that's the basal grade for troodontids, and birds are related but evolved serrationless teeth, long arms, dorsal ischial processes and such on their own branch.

The Jurassic Haplocheirus also supports this idea, since it shows serrated teeth and a general morphology so primitive that Cau's Sumrukia matrix found it to clade with compsognathids.  Note therizinosaurs also have mesial and distal serrations, and that Falcarius has made the clade more basal in most matrices, unlike earlier ideas they were related to oviraptorosaurs.  Maybe coelurosaurs were of the compsognathid-coelurid grade all through their evolution, with ornithomimosaurs, alvarezsaurids, therizinosaurs, dromaeosaurids and troodontids+birds each developing their birdlike and/or herbivorous characters separately.  This idea is kind of anti-Paulian or anti-BCF in nature and has plenty of precedent in the literature.  Ostrom long suggested Ornitholestes as a dromaeosaurid ancestor, and Makovicky (1995) found the two to be sister taxa to the exclusion of birds based on vertebral characters. 

Of course the real test is with cladistic analyses, so how does the idea fare?  The in progress Lori matrix finds a fairly traditional tree with Ornitholestes sister to Maniraptoriformes, microraptorians and Unenlagia basal among dromaeosaurids, and is somewhat unusual in finding troodontids sister to birds.   Constraining Ornitholestes to be a dromaeosaurid and the dromaeosaurid topology 'backwards' from the consensus (Achillobator,Dromaeosaurus(Deinonychus,Velociraptor(Microraptor,Sinornithosaurus))) results in trees 11 steps longer.  Not too bad when you consider enforcing Longrich and Currie's (2009) plausible-seeming traditional dromaeosaurid topology is 7 steps longer than the minimum.