Look for either an extra header row, which would treat a second row of header text as data and would be character or look for another row with some character other than digits, decimals and % sign. Maybe a dash or something used to indicate "no value" other than blank. Or you'll see the Excel indicator of "text", the little green triangle currently that appears in one corner of the cell. That indicates someone typed an ' before the value and is now text.
Save as to CSV is a good thought but did you look at the TEXT resulting? As in with a text viewer like Notepad? Typically a bit hard to read but should see likely see something in the first 20 or so rows of data. Look for quotes around the percent field, possibly with one or more leading spaces inside the quotes if no other odd characters.
Import + Excel = guessing game.
Save Excel to CSV and read with a data step will control variable types, names and properties and likely through an invalid data message or more when something odd happens to the Excel source.
One of my favorite reading Excel files was more interesting as part way through the file columns that had been dates started appearing as currency (none of the SAS date informat like currency values), the income column started appearing as dates (the Comma informat typically used to read currency doesn't like yyyy-mm-dd values).
For that particular issue since the cell properties were the problem the solution was highlight the entire column and set the value type as all dates or all currency ( and a similar fix had to be done to a couple of Id number columns that were partially currency). And this was from a source that was computer generating the files (with a long history of stupidity). Human entered values are worse.
