From the documentation:

STG_LOSS: Coefficient that represents the annual storage losses of a storage process p in region r, as a fraction of the (average) amount stored, corresponding to a storage time of one year. If the value specified is negative, the corresponding annual losses are interpreted as an annual equilibrium loss (under exponential decay).

Thus, the storage loss is proportional to the time the amount of energy is stored in the storage. For a DAYNITE storage, the storage time is typically only a few hours, and therefore the annual loss of 0.2 would be very small indeed: with a 10 hours average storage time the loss would be only 0.02%, so hardly visible at all. If you want 20% loss per day, you could try multiplying 0.2 × 365. However, if the losses are very high, they might easily even exceed the storage level, and therefore one might recommend using the equilibrium loss approach instead when the losses are very high. So, for 20% hourly loss I would recommend the negative value: −0.2 × 8760.

STG_LOSS: Coefficient that represents the annual storage losses of a storage process p in region r, as a fraction of the (average) amount stored, corresponding to a storage time of one year. If the value specified is negative, the corresponding annual losses are interpreted as an annual equilibrium loss (under exponential decay).

Thus, the storage loss is proportional to the time the amount of energy is stored in the storage. For a DAYNITE storage, the storage time is typically only a few hours, and therefore the annual loss of 0.2 would be very small indeed: with a 10 hours average storage time the loss would be only 0.02%, so hardly visible at all. If you want 20% loss per day, you could try multiplying 0.2 × 365. However, if the losses are very high, they might easily even exceed the storage level, and therefore one might recommend using the equilibrium loss approach instead when the losses are very high. So, for 20% hourly loss I would recommend the negative value: −0.2 × 8760.