(1) This is an excellent site you found...
http://jdebp.info/FGA/bootstrap-programs.html
Acc. to Wikipedia, partition Type/ID of 0x0A is an OS/2 Boot Manager partition (BMP). Because you have the partition, yours indeed is the IBM Boot Manager (not the Partition Magic version which lives in the MBR). Your bootstrap is...
(a) Firmware (BIOS) loads the MBR (of BMP the target drive).
(b) The code in the MBR searches its partition table for the Active partition (which is BMP).
(c) The code then loads the VBR (Volume Boot Record) of BMP.
(d) The VBR code loads the IBM Boot Manager.
(e) Finally, it says...
"IBM Boot Manager itself then reads the partition tables from all (firmware-accessible) non-removable DASDs [direct access storage devices], and presents a menu to the user of all partitions that it finds that have been marked with the 'bootable' flag."
So -- since you got the menu (naming eCS & Windows), all went well that far. BIOS did its job. The MBR did its job. The VBR did its job. Then, you report...
"Selecting either the screen goes black for a second and then reboots to the BIOS screen."
The IBM Boot Manager presented the right partitions to choose from. It found the "extended flags" that say "bootable" in the partition table. You made the choice. But the manager then failed to load the OS of the partition you chose. It could mean the dimensions of the partitions that also are in the table have been mussed -- but the dimensions of BMP had to be good enough for steps (c) & (d) above.
So, maybe it's something else gone wrong. That excellent article doesn't say in detail exactly what happens to load the OS after you make your selection. We can see in your BCD that the Windows Boot Manager would pass control to Winload.exe. But what does the IBM Boot Manager pass control to?
(2) "But making Win 10 startable and active and trying to boot gives the 'No physical...' message.
Not sure how the IBM Boot Manager boot problem and the Win 10 by itself boot problem are related."
(a) I think "startable" & Active are the same thing. DFsee does the same as DiskPart would do to make the partition Active. Did it automatically turn off Startable for the eCS partition? (DiskPart wouldn't). You can have only one Active partition per HDD. Edit: But Startable (I think) also means that the partition's VBR (volume boot record) has been changed to load the IBM Boot Manager.
(b) It is "bootable" that DiskPart can't do. Some kind of "extension" has been added to each entry in the partition table for that. And it looks like code has been put into its VBR (volume boot record). That's what makes step (d) work above. Edit: Sheesh, that's in the Startable's.
And, later at the site, it says, "IBM's boot manager code is stored as a straight sector-by-sector image right after the VBR, and the [Boot Manager] partition is not formatted as a regular disc volume, so the boot loader code in its VBR is somewhat simpler [than Microsoft's bootmgr]."
(c) Therefore, when the Windows partition is made Bootable, I think it is still the IBM Boot Manager that boots it. I wonder whether we should try to do an installation like at that os2World site I found that uses the Windows Boot Manager?
(3) Another possible location for WinRE.wim is the System Reserved partition. But yours didn't show the Recovery\WindowsRE folder & is disabled anyway. Your BCD shows the recovery environment is not enabled. And there is no recoverysequence entry that would set it up. So, you must always boot to external media for yours. That's ok. We'd better let it alone than give eCS another reason to fail. Otherwise, that BCD is good. It knows where the boot files & OS partition are. But it is a single-boot BCD. It is not being used to do the boot.