Maybe you simply need to switch off the storage of segmentation information (mrk tags) in the target file? Go to project settings, file types and check in your XLF file type, if the option "Add segmentation information to the translated file" (or similar) is activated. If so, uncheck and save target file again.

Thanks for the speedy response, Jerzy. That box is already unchecked.

Yes, in File -> Options. Please check this in your project settings, as I stated above, NOT in File -> Options.

And the target file being double the size of the source file is fully ok, as it contains source AND target text, while the source file might have just contained the source text.

Hi Jerzy, I have the same settings under "File > Options > File Types > XLIFF" and "Project Settings > File Types > XLIFF".

I see what you mean about the file size (d'oh!).

I'm waiting to hear back from the customer about which tags have been duplicated, although I don't know much about that side of things.

My wild guess is that the customer has actually little to no knowledge about the XLIFF format and therefore sees things like the target segments or other perfectly fine XLIFF stuff as "duplications".

Let's see how close to the reality this guess will be...

Hi Jerzy, so it turns out there are indeed a few anomalies in tag text after saving as target. 

The tag text: "></g>< has been replaced with: " />< a few times. 

Similarly, the tag text &#xD; has been replaced with a hard return.

These changes don't seem to be problematic. What is definitely problematic, however, is that Trados has added a load of tag text to the source text (perhaps also the target, but I haven't checked that yet) in one translation unit.

Do you have any idea why that would happen?

It looks like the forum has eaten some parts of your message - some of the pointy brackets were understood as HTML tags...
You would need to format the tags as <code>... which is unfortunately very tricky as the forum editor does not offer the same full formatting functionality in replies as when creating new post :(

How strange, it looks fine to me - here's a screen grab:

Ah, well, okay, this is what I see as well... but it doesn't make much sense, so I thought there is something missing...

Include more context, i.e. more of the surrounding text. This is not enough to understand what is your issue.

Basically, tags with empty content (<foo></foo>) are syntactically identical to empty tag (<foo/>), so if you are wondering about this replacement, then it's NOT a problem.

Similarly, replacing the Line Feed numeric entity (&#xD) with Line Feed + Carriage Return makes no difference if made in the structure of XML itself.
It would be a problem ONLY if made inside the actual textual content of the XML elements... which I believe did not happen.

So, so far I do not see any problem from the XML/XLIFF perspective...

I'm also wondering about the claim that "Trados has added a load of tag text to the source text" - can you post some screenshots, examples, etc.?
I somehow don't believe it's actually anything significant from the translatable content perspective...

Hi Evzen, thanks for looking at this. Truth be told, all this programming language is Greek to me.

To try and track down the "duplicated tags" in my translated xlf file, I copied and pasted the whole of the source xlf file into one Word doc then the whole of the target xlf file into another Word doc. After doing that, I used the compare&merge function in MS Word to highlight the differences between the two files.

On the whole, everything looks as it should - the target xlf file contains the additional, translated target segments (additions are shown in pink):

However, in the case of one particular translation unit, Trados Studio seems to have inserted additional tag text into the source:

When the customer tries to import the translated xlf file, this is precisely the point at which the import fails.

Does it make any sense to you?