Cambridge, Oxford, Imperial, and Bath all have strong Maths programmes with signifcant Applied Maths/Theoretical Physics content and/or joint honours courses in these areas which build on these strengths. They also have excellent to good Physics departments. Edinburgh and Glasgow also both have very strong Physics courses (I'm not sure about maths) - the former being the home of the Higgs Centre and the current academic residence of it's namesake to my knowledge. Warwick has a very strong Maths course, however I think they have a touch less physics content in it - I'm not sure how their Physics department is specifically, however they are quite well regarded overall and I doubt it's "bad" by any means, and likely benefits from the strengths of it's Maths department to some extent.
Manchester, UCL, Sheffield, Nottingham, KCL, Birmingham and Southampton all have good undergraduate Physics programmes - the latter at the least has the opportunity for the best students to spend their final year at CERN or the Harvard-Smithsonian Observatory doing a yearlong research project. Another topic of note is that Birmingham has the option to "intercalate" a year in the Computer Science department - following the second year of their CS course I believe. This would be very useful for budding theoretical physicists (especially astrophysics) or experimentalists who anticipate a large amount of modelling work (condensed matter experimentalists for example) as preparation for their PhDs. Oxford also is the leader in the Philosophy of Physics, and has an excellent joint honours in this course. Unlike other such courses, this is essentially a theoretical physics degree, with some philosophy of physics and science - the latter replaces the experimental components. It covers all the essential mathematical and physical content you need, and their philosophy of physics courses necessarily assume all of this content and so it's not merely fluffy minded discussion of "quantum mechanics is weird". Indeed, fundamental questions about the nature and structure of quantum mechanics and space-time physics are very important for theoretical physicists, and continue to be relevant even today (for example the Bell inequalities). It's also worth noting Maths at Cambridge has the first year option of "Maths with Physics" which allows you to either continue in the Maths course as typical for second year onwards, or to switch into Natural Sciences Physics/Astrophysics (i.e. Experimental Physics - although the Astrophysics course does have some non-trivial overlap with the Maths course at the Part III level, and you can go into Part III Maths from Part II Physics/Astrophysics potentially).
However Oxford, Cambridge, Imperial and slightly further behind, Edinburgh are by far and away the frontrunners. For Theoretical Physicists especially, Maths at Cambridge might be one of the best choices for an undergraduate course - indeed many theoretical physicists from various locations pursue the Cambridge Part III Maths course as a standalone masters. Oxford however has recently (a year or two ago) launched a similar masters in Mathematical and Theoretical Physics. Imperial has it's well established QFFF course for theorists as well, although it also has a great number of more experimentally oriented masters courses (and options in it's undergrad course). If you were interested in having some exposure to the philosophical side of things (a fairly reasonable prospect for a potential theoretical physicist) it's quite common for the Oxford PhysPhil students to continue to Part III Maths at Cambridge and then onto PhDs in various places. This would certainly give you the "best of both worlds" if this is your aim. If you're more on the experimental side, this may not be as appealing an option though.
Additionally as touched on above by other posters, certain universities usually have specialisms in certain areas and have large and/or active research groups in these areas. This usually manifests in undergraduate courses as optional modules and possible 3rd/4th year projects/dissertations/theses. As a result, this might cause your to lean more to one than another. Manchester is of course the birthplace of graphene, so has a fairly well known graphene group. However Exeter also has quite a prolific graphene group - they're also known for Exoplanet research. UCL has a good amount of planetary science and space physics research ongoing, while Cambridge has quite a lot of condensed matter and materials physics work (sometimes working with their EE or materials groups/departments) and fundamental physics (especially through maths). Southampton has a reasonably sized gravitation group in their maths department which has a number of links to their physics course, as well as a great deal of optoelectronics work between the physics and EE departments. Imperial, as well as a fair amount of theoretical work and applied maths, has some particular expertise in fluid/continuum dynamics and shock physics as well as quantum computing and optoelectronics. As a result, one university which may be considered "worse" by a league table, or even my comments, may be better for you.