The returned tenders can't be compared

Three numbers are on the table and each one answers its own version of the enquiry: different exclusions, different attendances assumed, design risk carried in one, struck out of another, unmentioned in the third. The spread between them is not the market speaking, it is three different readings of the same package.

Picking the lowest raw number buys the biggest set of gaps, because the cheapest tender is usually the one carrying the least scope, and the difference arrives later, priced as variations at rates you no longer control. Levelling is measurement work, not admin: every return pulled onto one schedule, the exclusions priced, the assumptions stated, and only then the comparison made.