The CURSORTOXML() considers Blob, Memo/Binary and varbinary to be a binary data and uses BASE64 encoding on them. The BASE64 encoding will increase the size by about 30%.
The Memo is a character field and is included in XML w/o changes. A binary data in a memo will cause XML to become invalid.
You cannot expect 28MB of data to be compressed into 4KB. Most likely 4KB file was truncated.

>After the SWFOX sessions, I have been using Craig Boyd's compression FLL and Alan Stevens suggestions on using to compress XML files. In testing so far, I attempted to put into a memo file a compressed XML. However on trying to get the cursor back after compression I kept getting the following error.
>
>

"XML Parse error: An invalid character was found in text content."

>
>So instead of a memo field, it made it a Blob field and was successful in getting the cursor back with out errors.
>
>My question is, why does this seem to work and secondly is there any possible issues that I am not thinking of by using the Blob field?
>
>(by using the Blob field, I am able to compress 28MB of data down to 2MB. I was real excited when using the memo field i was able to compress down to 4kb, but could not extract.)