A record sample is unbiased because it treats data as a stream of equally valued records, using a nominal value of 1, regardless of the actual values in a record. This means that each record has an equal chance of being selected. The sampling population is therefore the number of records in the file. Consequently, a $1,000 item and a $1 item have an equal chance of being selected, since each item is only one record and each record has an equal chance of selection. So, there is a significant probability that very large transactions will be overlooked.
Record sampling is most useful for compliance or understatement testing. In compliance testing, you are more concerned with the rate of errors in the total population. In understatement tests, you are more concerned with items that are missing or recorded at too small a value.
It may be that the larger amounts are least likely to be understated as they may be subjected to extra controls not present for smaller amounts. Choosing a sampling method that biases large amounts may result in missing a potential problem relating to smaller transactions.