Forecasting future point with partial future data already available

Working on a forecast model that should output an End of monthly value, the interesting part is that we already have partial (90%) of that data available at the prediction point (max 30 days away).

The purpose is to take into account the current monthly trend and project that out until the end of the month, but also take into consideration that we already know those future points with 90% confidence.

For example:

Let's say we're in the 3rd day of the month:

Day 1: 10 in sales

Day 2: 15 sales

Day 3 (Current): 20 sales

Day 4 (Future): 15$ sales with 90% confidence

Day 5...29 (Future)

Day 30 (Future): 20$ sales Total: 600 in sales (45 current, 550 future)

How do we forecast that 600 in sales?

I've tried Multiple Linear Regression with the X taking into account lagged metrics and current daily, with the Y as the true value as 600, to essentially force the model to always predict t+1 as end of month value, but this is not optimal obviously. This was also done with various Lasso + Ridge penalties.

Surprisingly a Random Forest Regressor performs almost to well, in fact it overfits perfectly :)

I have the feeling this could be stated as some probalistic Baysian model, but not sure on what to search or look for.

Any ideas would be greatly appreciated!