An error count metric provided to Datadog from Cloudwatch is a number of errors per minute. Because all Datadog metrics are stored at a 1s resolution, we divide this by 60 during our post-processing.
To get closer to the value you are expecting, you can append
.as_count() * 60 to your query. Note that you may not see the exact numbers you see on Cloudwatch, because Cloudwatch only makes data available every 5-10 minutes, and our AWS crawler runs every 5 minutes.