@dinhe878 your table/files are very useful and give the reason clearly: the labels need to match at each rank tested. Imprecise matching is not allowed, and this includes rank labels (e.g., "g__"). So this means that the exact same reference database version must be used. It looks like the rank label "k__" is used in the expected, but "d__" in the observed sequences. So you must have used a different reference database for each, but you could also manually fix the labels in your "expected taxa" file so that they use the same naming conventions.
Good luck!