Why doesn't eMILPO and iPERMS backup off each other?

They are two different systems that were developed to replace to antiquated systems, at the time, no thought was given to merging them.
iPERMS was developed in 2004 to replace the old system of mailing records to Indianapolis to get into your microfiche record.
eMILPO was developed in 2003 to replace SIDPERS3, a DOS-based program that had to run in a JRE emulator after Windows 3.1 was fielded in the Army. (the Guard still uses SIDPERS in an Oracle database environment, running in a JRE).
iPERMS allows the user to upload and index a document. Almost nothing in the index fields is of any value to update a data point in eMILPO.
For example, you get a DA 1059 for completion of BLC. The indexing information in iPERMS is your SSN, (that routes it to your record), the document name (DA 1059 in this case) and effective date (date of graduation for this one). There is nothing there to tell eMILPO in any kind of data cull that you completed BLC. For non-1059-producing schools, the name of the document is Crs Compl 40 hrs or crs compl less than 40, that does not tell you what school it is at all. Indexing an MOS or ASI order (ORD MOS ASI) does not give the MOS or ASI in the indexing, so also worthless.
There are limited cases where a data cull could work, like a DA 67 or DA 2166 effective date is the thru date of that OER and NCOER, and that would be a handy update from iPERMS to eMILPO, but that would require a new contract with the vendors that code iPERMS and eMILPO.
Another even more expensive way would be to rewrite eMILPO so that it becomes the document loading feed to your OMPF, with the indexing screen providing additional indexing fields that update your eMILPO.
For now the process is:
1. a source document goes to your iPERMS.
2. an update is coded in eMILPO in reference to a source doc in your record.
A source document can make it to your record without someone making the eMILPO update, but there should never be an update to your eMILPO without a source doc at least on its way to your record in an iPERMS batch.

Doubt worry, if IPPS-A ever launches you'll have a brand new system that cost millions to develop with the revolutionary idea to merge all the systems only to continue to maintain all of them as redundancies!

Because the Army hires different contractors to build out its systems, and interoperability with older systems is seldom a consideration. Emilpo was supposed to tie all this stuff together. So was DIMHRS. Probably others.
It is the bane of HR guys existence.

The two systems do not feed off of each other. What should happen is this; when the S1 update the SRB using eMILPO, they also suppose to scan the documents to upload in iPERMS. So if you’re seeing the update in one but not in the other, that means S1 didn’t do both transactions as required.

Now if you’re at a 1059 producing school, the school is required to make sure HRC receive a copy because the S1 cannot do the update on 1059 producing schools... I hope this helps...

Separate contracts, separate vendors.

Iperms is just a document storage. It doesn't read any of the data on the document. For instance, you finish BLC and a 1059 is uploaded, or you receive an award and it's uploaded. Iperms can't automatically read that award to see what it is or that 1059 to see if you pass.
Emilpo is more like an index or table of contents. Iperms holds the source document and emilpo is just an index of your assignments, awards, and certifications. Emilpo can't generate the award or orders simply by being updated to reflect you have a new award.

Your S1 is the human link to the two. What should happen is S1 takes your source document, updates emilpo and then uploads the source document as proof for the change. This online repository is used by you when you lose everything and by HRC when they audit financial reviews.