约8%日期时间字符串转换年月错误，求技术排查解决

我需要计算起止日期之间的时长，但日期格式不固定。我编写了字符串清理和多格式解析函数，但约8%的日期转换不准确，出现119条结束日期早于开始日期的错误记录，无法定位问题。

测试数据包含多种格式的起止日期，错误转换的记录也已整理出来，以下是我的代码：

data <- read.csv('justDates.csv',
                 na.strings=c("","NA"))


## FUNCTIONS -----------------------------------------

# clean text
# don't judge me, I hate regex and this is hella ugly but gets the job done. Maybe. 
clean_dates <- function(datetime_string) {
  # Normalize timezones by removing all mentions of timezone abbreviations except UTC
  datetime_string2 <- gsub("\\s*\\([^\\)]+\\)", "", datetime_string)  # Remove extra timezone notes in parentheses
  datetime_string3 <- gsub("(?<=\\d)(?=UTC|\\sUTC).*", "", datetime_string2, perl = TRUE)  # Strip trailing text after UTC
  datetime_string4 <- gsub("[\n\t]", " ", datetime_string3, perl = TRUE)  # Remove newlines and tabs
  datetime_string5 <- gsub("UTC\\s+", "UTC", datetime_string4, perl = TRUE)  # Clean trailing whitespace after UTC
  datetime_string6 <- gsub(" ?(UTC|GMT|UT)", "", datetime_string5, perl = TRUE)
  datetime_string7 <- gsub("[\\h\\v]+$","",datetime_string6, perl=TRUE)
  return(datetime_string7)
  
}

#convert various date and time formats into one standard format
parse_date <- function(date_string){
  
  parse_attempts <- list(
    function(x) mdy_hms(x),
    function(x) dmy_hms(x),
    function(x) mdy_hm(x),
    function(x) dmy_hm(x),
    function(x) ymd_hms(x),
    function(x) ymd_hm(x)
  )

  # # Attempt each format, catching errors and returning the first successful conversion
  for (parse_fn in parse_attempts) {
    date <- tryCatch(parse_fn(date_string), error = function(e) NA)
    if (!is.na(date)) return(as.numeric(as.POSIXct(date))) # Convert to consistent POSIXct format if successful
  }

  return(NA) # Return NA if no formats matched
  # Attempt to parse the date with each format
  
}
  

  
  
# Clean the Dates
df_dates <- data |>
  mutate(
    start_date_clean=sapply(start_date,clean_dates),
    end_date_clean=sapply(end_date,clean_dates)
  )

# Convert to POSIXct
df_dates_posix <- df_dates |>
  mutate(
    start_datePOSIX = sapply(start_date_clean, parse_date),
    end_datePOSIX= sapply(end_date_clean, parse_date)
  )

# Check dates are correct
df_dates_comp <- df_dates_posix |>
  mutate(
    neg=end_datePOSIX<=start_datePOSIX,
    start_date_rev=as_datetime(start_datePOSIX),
    end_date_rev=as_datetime(end_datePOSIX)
  )

test3 |>
  filter(neg==TRUE) |>
  nrow()

经检查有119条记录的end_date早于start_date，对比*_date_rev与*_date_clean列可发现转换错误，但无法定位原因。

恳请各位提供帮助与见解，我哪里出错了？

内容的提问来源于stack exchange，提问作者zaneywolf