You can use
(?<!\d)\d{1,2}/\d{1,2}/\d{2}(?:\d{2})?(?!\d)|\b(?:charB|wordA)\s*\d*[.,]?\d+|(?<!\d[.,])(?<!\d)(\d*[.,]?\d+)(?!\s*(?:wordB|charA)|[.,]?\d)
Get only those matches that are captured into capturing group #1. See the regex demo. Details:
(?<!\d)\d{1,2}/\d{1,2}/\d{2}(?:\d{2})?(?!\d)| - a date-like string: no digit allowed immediately on the left, then one or two digits, /, one or two digits, /, and then two or four digits with no extra digit on the right allowed, or\b(?:charB|wordA)\s*\d*[.,]?\d+ - a word boundary, then charB or wordA, zero or more whitespaces, zero or more digits, an optional dot or comma, one or more digits| - or (the next part is captured, and re.findall will only output those in the resulting list, the above ones will be discarded)(?<!\d[.,])(?<!\d)(\d*[.,]?\d+)(?!\s*(?:wordB|charA)|[.,]?\d) - no digit or digit and a . or , allowed immediately on the left, then zero or more digits, an optional . or , and one or more digits are captured into Group 1, and then the negative lookahead fails the match if there is wordB, charA or an optional . or , and a digit appear immediately on the right after any zero or more whitespaces.
See the Python demo:
import re
text = '4.5 $55 1,200 wordA 3 sometext 2 wordB sometext 4.3charA sometext charB21.6 sometext 11/10/22'
rx = r'(?<!\d)\d{1,2}/\d{1,2}/\d{2}(?:\d{2})?(?!\d)|\b(?:charB|wordA)\s*\d*[.,]?\d+|(?<!\d[.,])(?<!\d)(\d*[.,]?\d+)(?!\s*(?:wordB|charA)|[.,]?\d)'
matches = re.findall(rx, text)
print( [ m for m in matches if m ] )
# => ['4.5', '55', '1,200']
