Quantlib Vanilla Swap Amount not based on Forwards

I have the following code: cashflows = pd.DataFrame({ 'Nominal': cf.nominal(), 'AccrualStartDate': cf.accrualStartDate().ISO(), 'AccrualEndDate': cf.accrualEndDate().ISO(), 'AccrualPeriod': cf.accrualPeriod(), 'Price': cf.price(disco_yts), 'Rate': cf.rate(), 'Amount': cf.amount(), 'Forward': cf.indexFixing(), 'FloatAccruedAmount': cf.accruedAmount(valuation_date+1) } for cf in map(ql.as_floating_rate_coupon, swap.leg(1)))

    fixed_cashflows = pd.DataFrame({
        'FixedAmount': cf.amount(),
        'FixedAccruedAmount': cf.accruedAmount(valuation_date+1)
    } for cf in map(ql.as_fixed_rate_coupon, swap.leg(0)))

swap.NPV() My issue is that amount (from floating leg) is not as expected. Mathematically I want amount to be amount = nominal x accrualperiod x forward , but instead it seems as though Quantlib is doing something completely different? This is strange because 99% of the time it does create amounts as expected. What am I missing? TradeId Nominal AccrualPeriod Price Forward Amount ExpectedAmount 0 bob -2421350000 0.084931507 -20443249.84 0.06458 - 20 478 517.87 -  13 280 806.23 1 bob -2421350000 0.252054795 -61382112.76 0.067465903 - 62 536 259.86 -  41 175 309.17 2 bob -2421350000 0.243835616 -63886355.38 0.077314926 - 66 312 009.20 -  45 647 611.25 3 bob -2421350000 0.252054795 -65823488.94 0.079179836 - 69 685 424.07 -  48 324 473.39 4 bob -2421350000 0.252054795 -65385660.54 0.080726366 - 70 629 291.20 -  49 268 340.52