Editor prevents authors from entering a handful of reserved characters, ampersand or & being one of them, by escaping them and converting them into their character entities. There are a number of ways to bypass this. One is to use the Insert Symbol menu/button/dialog. Look for the omega button on the ribbon or Symbol on the Insert menu. Once the dialog is open you may have to switch fonts and/or navigate through the selected font.
If you like to use the keyboard and you're going to insert one or two entities lots and lots, you could map a keyboard shortcut to that or those entities. Look in help under "map"
You can also edit a bit of content as XML Source (look on the Edit menu near the bottom). In that dialog, you can type any characters you like and they will be inserted as-is. Careful there, though, you schema/dtd is not enforced and what you create could cause problems when finally inserted into your main document. I think you get errors (and may not even get an option to insert) invalid markup. (I don't have Editor on the machine I'm using so I can't validate the details of some of this.)
In all cases, once you have □ inserted correctly, your charent.cf file will need to be properly set up. If it is not, you may not get the output you require. Also, you may need to pay attention to how you are including fonts (assuming you are outputting PDF).
Finally, be warned that if you are using any Arbortext technology that will cause your output to go through the content pipeline (profiling, for example, will do this, DLM will do this), your charent.cf settings may not be applied at the appropriate time. I ran into this issue a long time ago, so it is also possible that this has been resolved in newer versions. But my charent.cf was firing on the first parse in the pipeline at a time when the required fonts were not available so our output was not correct. Arbortext, at the time, did not consider this a bug, and offered a DTD/stylesheet workaround which we have used since. We created elements for specific fonts and created attribute values to choose specific characters. For example (again from memory, so the character / font may not match up in real life, but the strategy is valid), <arial_ms type="euro">. Using this strategy the element is preserved until the final stage of the content pipeline which is where the stylesheet is applied.
Also, you should check out the mailing list "adepters" which is where lots of Arbortext users hang out and provide peer-to-peer support. Info on that mailing list can be found here: