i didnt have time to verify or reproduce or check the things so i cant reply any queries
i will just post generic thoughts on certain highlights

mft is was always 2 fragments as far as i have seen
if i have a severely fragmented mft i run fsutil usn deletejournal /n <drive> and recreate it (use query and save the max and alloc numbers first)
this normally brings back mft to 2 fragments

i normally run some boot defraggers (ultradefrag from source forge for example ) windows defrag yells that it needs 15% commision to work i cant even afford 1 % sometimes on my real machine

at the moment my mft is 93% full with 2 fragments and the fragmented file is always $bitmap

Master File Table (MFT) fragmentation
    Total MFT size                             = 308 MB
    MFT record count                           = 295,224
    Percent MFT in use                         = 93 %
    Total MFT fragments                        = 2

Fragments 	Size 	Filename 	         Comment 	Status
10	       40 Kb	C:\$MFT::$BITMAP	- 	locked
as far as i have seen defragging BITMAP is kinda impossible as it has details about each and every sector including a reference to itself

i still dont believe shell and ole ole plays significant role in the process the object is created in R0 with ObCreateObjName & sisters and the handle is created with obpCreateObHandle and brothers does shell and ole send some undocumented DeviceIoControl to ntfs.sys

NtCreateFile ->IoCreateFile -> IopCreateFile->ObopenObjectByName -> lookup which goes on into obinsertDirectoryEntry etc loops is how a new file is created