Header row(s) not extracting properly

[bhoppeadoy]

I'm trying to extract the table in this PDF: ads7828-4.pdf

Here's what I'm doing:

from gmft.auto import AutoTableDetector, AutoTableFormatter
from gmft.auto import AutoFormatConfig, AutoTableFormatter
from gmft.pdf_bindings import PyPDFium2Document

detector = AutoTableDetector()
config = AutoFormatConfig()
config.verbosity = 3
config.semantic_spanning_cells = True
#config.enable_multi_header = True
#config.large_table_assumption = True
formatter = AutoTableFormatter(config)

doc = PyPDFium2Document("ads7828-4.pdf")

tables = []
for page in doc:
    tables += detector.extract(page)

from IPython.display import display

ft = formatter.extract(tables[0])
display(ft.visualize(filter=[5]))
f = ft.df()

print("\nTable Headers:")
print(df.columns.tolist())

Visualizing the header row, looks like its capturing several extra empty cells.

And so the headers come out like so:

What I want is for the headers to come out like this:

I'm not sure what I can tweak to clean up how it detects the header cells. Any suggestions?