Playing with Forecasting/Analyse Models

Hello all!

First time poster here

I have had a quick look on the forums for some threads for people using forecasting or "Analyse" in analytics but haven't been able to find anything that gives me confidence I'm doing this correctly... So here is my first attempt below. This is a straightforward widget with daily ticket sales counts for one of our venues going back right to the beginning of when we went live. In other words it has plenty of data feeding it and it's not too complicated in principle. 

What I am trying to achieve is to come up with a forecasted daily total for total ticket sales based on previous years of data. 

I have managed mostly to only get flat lines and in some cases wildly inaccurate results. Now, manipulating the parameters until I get something that looks even remotely close I have managed the below.

In the below image, the gold line are our business's existing daily forecasts based on our own management accounting. I wanted to try to get the forecasting model to come as close as possible to this (obviously with an open mind that some of our human forecasting might be wrong). The only way I can manage this is to switch to Prophet model, exclude "last ~30 days of data" (because it appears to be reading the data as though existing sales for future performance dates are valid sum totals for those days, and even then, I would say it's only the lower bound that is roughly following the expected peaks and troughs I think will actually occur.

Obviously this is not a way forward as we wouldn't use this for anything reasonable unless I had confidence in the methodology behind the forecasting model... but I'm just keen to know if anyone has any advice of what potential next steps to take to try to understand this better? 

Here are some adjustments I've already made:

1) Reducing the forecasting period down so it's less than 10% of the total data set feeding it

2) Adding a restriction so it can't produce values <0

3) Changing the forecasting models to different types

4) Changing the confidence percentage

5) Excluding last "x" days of data. 

I have a hunch that there is some logic problem here related to the fact that this is fundamentally SALES data but I've got it split by PERFORMANCE date (because that is how we have to do our targets). Can anyone help?

Here's the chart, again green is actual sales, gold is hardcoded numbers which show our external-to-tessitura daily targets, and the pale green is the forecasting model. The full chart has daily data from Nov 2020 onwards. 

Thanks in advance for any advice