That's a fascinating discovery.
I think the Seek lenses are generally made of moulded chalcogenide glasses, mainly I believe for cost reasons (a moulded glass lens is way cheaper than a diamond-turned germanium lens, even if the glass itself is fairly pricey). Schott, for example, offer various glasses with transmission between about 0.7µm (just-visible deep red) all the way out to >15µm:
So, yes, it does look as though there's a prospect of some transmission through the optical system (though it's unlikely to operate particularly well outside its design frequency range - the lens components will all be of sub-optimal configuration, leading to intrinsically blurry images). The coatings on the lenses (if any?) will be what really determine their out-of-band properties.
But you're right, a suitable filter should mean you can limit the passband to wavelengths other than the original design, though at a (significant?) performance penalty. How well the FPA actually responds to those other wavelengths is also an open question.
(You may be aware that at the other end of the spectrum some conventional digital cameras have, or can be modified to have, a response from about 1100nm to about 350nm, near-IR to near-UV, and that some supposedly visible light only lenses can perform surprisingly well outside their design bandwidths. Below is a sequence of images in near-IR, visible, and near-UV taken with regular lenses and suitable filters on a Fujifilm IS PRO, a long-discontinued 'full-spectrum' camera).
For comparison, here's a very poor thermal image of a different flower vs its visible light equivalent.
Thermal images of flowers are challenging, unless you can find a warm-blooded plant.
You may have discovered the first example of a comparable MWIR-LWIR imaging setup.